Publish the 2026Q3 edition of yolov8-seg

.gitattributes

@@ -42,3 +42,6 @@ hailo/yolov8s-seg-int8.hailo8l.hef filter=lfs diff=lfs merge=lfs -text
 jetson/yolov8m-seg-fp16.engine filter=lfs diff=lfs merge=lfs -text
 jetson/yolov8n-seg-fp16.engine filter=lfs diff=lfs merge=lfs -text
 jetson/yolov8s-seg-fp16.engine filter=lfs diff=lfs merge=lfs -text
+jetson/yolov8m-seg-int8.engine filter=lfs diff=lfs merge=lfs -text
+jetson/yolov8n-seg-int8.engine filter=lfs diff=lfs merge=lfs -text
+jetson/yolov8s-seg-int8.engine filter=lfs diff=lfs merge=lfs -text

README.md

@@ -25,22 +25,22 @@ model-index:
         metrics:
           - name: "Box mAP@0.5 (Nano ONNX FP32)"
             type: map_50
-            value: 51.13
+            value: 51.43
           - name: "Mask mAP@0.5-0.95 (Nano ONNX FP32)"
             type: map
-            value: 29.14
+            value: 29.24
           - name: "Box mAP@0.5 (Small ONNX FP32)"
             type: map_50
-            value: 60.02
+            value: 60.32
           - name: "Mask mAP@0.5-0.95 (Small ONNX FP32)"
             type: map
-            value: 34.83
+            value: 35.03
           - name: "Box mAP@0.5 (Medium ONNX FP32)"
             type: map_50
-            value: 65.08
+            value: 65.52
           - name: "Mask mAP@0.5-0.95 (Medium ONNX FP32)"
             type: map
-            value: 38.14
+            value: 38.62
 ---
 
 # YOLOv8 Segmentation — EdgeFirst Model Zoo
@@ -56,7 +56,8 @@ YOLOv8 Segmentation models trained on [COCO 2017](https://edgefirst.studio/publi
 Part of the [EdgeFirst Model Zoo](https://huggingface.co/spaces/EdgeFirst/Models). New here? **[Introducing the EdgeFirst Model Zoo](https://huggingface.co/blog/EdgeFirst/model-zoo-intro)** covers what is in the zoo, why every figure links to its validation session, and how to reproduce any of it on your own hardware. Want a platform, model or task we don't cover yet? [Vote on what we measure next](https://github.com/orgs/EdgeFirstAI/discussions/categories/polls).
 
 > [!TIP]
-> **Training experiment**: [View on EdgeFirst Studio](https://edgefirst.studio/public/p-695/exp-885/trainings) — dataset, training configuration, metrics, and exported artifacts.
+> **Training experiment**: [View on EdgeFirst Studio](https://edgefirst.studio/public/p-695/exp-c1f/trainings) — dataset, training configuration, metrics, and exported artifacts.
+> Rows badged **2026 Q2** come from the [2026 Q2 training experiment](https://edgefirst.studio/public/p-695/exp-885/trainings), whose weights the repository's `2026Q2` tag holds.
 
 > [!NOTE]
 > Anchor-free DFL detection head. Detection and instance-segmentation variants.
@@ -69,9 +70,9 @@ Accuracy ceiling for each size, measured against COCO `val2017` (5,000 images) w
 
 | Size | Params | GFLOPs | Box mAP@0.5 | Box mAP@0.5-0.95 | Mask mAP@0.5 | Mask mAP@0.5-0.95 | Source |
 |------|--------|--------|-------------|------------------|--------------|-------------------|--------|
-| Nano | 3.2M | 8.9 | 51.13% | 36.02% | 48.09% | 29.14% | [v-e87](https://edgefirst.studio/public/validation/v-e87/details?mode=charts) |
-| Small | 11.2M | 28.8 | 60.02% | 43.87% | 56.46% | 34.83% | [v-e91](https://edgefirst.studio/public/validation/v-e91/details?mode=charts) |
-| Medium | 25.9M | 79.3 | 65.08% | 48.76% | 61.52% | 38.14% | [v-e8f](https://edgefirst.studio/public/validation/v-e8f/details?mode=charts) |
+| Nano | 3.2M | 8.9 | 51.43% | 36.23% | 48.23% | 29.24% | [v-1945](https://edgefirst.studio/public/validation/v-1945/details?mode=charts) |
+| Small | 11.2M | 28.8 | 60.32% | 44.19% | 56.81% | 35.03% | [v-194e](https://edgefirst.studio/public/validation/v-194e/details?mode=charts) |
+| Medium | 25.9M | 79.3 | 65.52% | 49.23% | 62.00% | 38.62% | [v-1951](https://edgefirst.studio/public/validation/v-1951/details?mode=charts) |
 | Large | 43.7M | 165.7 | — | — | — | — | — |
 | XLarge | 68.2M | 258.5 | — | — | — | — | — |
 
@@ -94,150 +95,176 @@ Row conventions in the table below:
 - **Rows whose `Δ` cell reads `ref`** are the float reference runs each quantized/compiled measurement is graded against.
 - **Rows without a number** under the metric columns are validation sessions currently in progress, or a session not yet linked to its ONNX FP32 reference. The Studio Source link tracks the current status.
 - **Rows whose `Δ vs FP32` cell carries a ⚠** are below our accuracy expectations for that platform (more than 10 percentage points under the float reference). The numbers are real measurements on real hardware, reproducible from the linked Studio session, and we publish them as-is; we are investigating the results to make improvements, and the next snapshot of this card will reflect any recovered accuracy.
+- **Edition badges.** A Source cell such as `v-XXXX · 2026 Q2` marks a result carried from an earlier EdgeFirst Perception Index edition: this configuration was not re-measured on the current weights, so the row keeps the measurement, Studio session and model artifact of the edition that produced it. Rows without a badge come from the current edition.
 - **Precision** varies by target: the ONNX reference rows are FP32; macOS CoreML and NVIDIA Jetson TensorRT run FP16; the NXP i.MX 8M Plus, NXP i.MX 95 Neutron, and Hailo NPUs run INT8. The NXP Ara240 DNPU runs a **mixed INT8/INT16** scheme — most of the model is INT8, with the box-regression path (and the ops feeding it) promoted to INT16 to improve localization accuracy.
 - **Decoder variants.** EdgeFirst ships three INT8 split-decoders — **smart**, **logical**, and **combined** (described under *Decode* below). Every one we have measured is listed; none is headlined over another, because which one wins is a property of the target, not of the variant. Full converter documentation: [EdgeFirst model conversion](https://doc.edgefirst.ai/latest/models/conversion/) — these are the converters used by this Model Zoo and the EdgeFirst Performance Index report.
 - **Runtime.** Which inference stack executed the model. It is a column of its own because one accelerator can be reached by more than one: on Qualcomm Hexagon HTP, `ONNX · QNN EP` is an ONNX Runtime EPContext binary while `LiteRT` is the LiteRT + QNN delegate path, and the two produce measurably different pipelines from the same weights on the same silicon. Compare rows that share a size, platform, precision, and decode to read that difference directly.
 - **Decode.** What the exported graph emits, and therefore what your application has to do with it. **Logical** keeps the head's outputs separate and named — boxes, scores, and for segmentation the mask prototypes and coefficients — so reading them needs no decoder beyond taking the tensors as they come. **Smart** cuts the graph further upstream, at a point chosen by dynamic range and sliceability: the operations that quantize worst are lifted out of the model and run outside it, which is what recovers the accuracy, and which is why it needs a decoder implementing that cut (in EdgeFirst deployments, the HAL provides it). **Combined** fuses the decode into the quantized graph — the standard upstream export. It is the least accurate of the three by construction, and the EdgeFirst converters do not emit it for quantized targets; rows appear only where one was measured. Compare rows sharing a size and platform to see what the cut buys on your target — on detection it is often small, on quantized segmentation masks it is usually decisive. A `—` means the run carries no quantized decode split at all (float ONNX / TensorRT / CoreML).
 - **Rows are not ranked.** Where a size and platform appear more than once, the rows are competing configurations — a different runtime, precision, decode variant, pipeline mode, or board — not duplicates. Only genuinely repeated measurements of the *same* configuration are collapsed, newest first.
-- **Platform-label suffixes.** `(FRDM)` / `(Phytec)` name the NXP i.MX 95 development board a session ran on. `— latency` / `— throughput` mark the two pipeline configurations the NXP i.MX 95 Neutron and NXP Ara240 targets run: the **latency** pipeline runs inference serially for the lowest per-frame latency; the **throughput** pipeline runs multiple inference workers for the highest FPS, which raises per-call inference time in exchange. Rows with neither suffix run a single pipeline.
-- **End-to-end (ms)** is the sequential per-image latency of the compute pipeline — preprocess → inference → postprocess. Image acquisition (camera or file load + JPEG decode) overlaps these stages and is excluded from this figure.
-- **Realized FPS vs Core-throughput ceiling (FPS).** **Realized FPS** is the *measured* steady-state throughput — the rate at which final results are actually delivered over the full validation pipeline. It normally exceeds `1000 / end-to-end` because the runtime overlaps stages across frames, and it is the **true, priority number**. **Core-throughput ceiling (FPS)** (shown with a `~`) is the accelerator's **core** ceiling — `1000 / device-compute-time`, the rate the NPU/DNPU could sustain *if it were the only bottleneck* — so it is a **possibly-achievable note, not a claim**. It is read from the isolated device-compute stage, which (unlike the host capture/preprocess stages, whose measured time inflates when the pipeline is backpressured) is stable and load-independent. Whether a deployment approaches it depends on the surrounding pipeline, and two levers dominate: (1) **host bottlenecks** — these validation runs decode a JPEG per image, whereas a live **camera pipeline** skips that decode and can run closer to the ceiling; and (2) **confidence threshold** — validation runs at `0.001` to capture every detection for mAP, which makes NMS/decode heavy, while a deployment threshold of `0.25–0.75` produces far fewer candidate boxes and lighter postprocessing, raising realized FPS toward the ceiling.
-
-| Size | Platform | Runtime | Decode | Box mAP@0.5 | Mask mAP@0.5-0.95 | Δ mask vs FP32 (pp) | Inference (ms) | End-to-end (ms) | Realized FPS | Core-throughput ceiling (FPS) | Source |
-|------|----------|---------|--------|-------------|-------------------|---------------------|----------------|-----------------|--------------|-------------------------------|--------|
-| Nano | ONNX FP32 (AWS Graviton · 4-core) | ONNX | — | 51.13% | 29.15% | +0.01 | 387.44 | 425.94 | 10.0 | ~3 | [v-e5d](https://edgefirst.studio/public/validation/v-e5d/details?mode=charts) |
-| Nano | ONNX FP16 (AWS Graviton) | ONNX | — | 50.29% | 28.67% | -0.47 | 277.88 | 360.67 | 6.9 | ~7 | [v-711](https://edgefirst.studio/public/validation/v-711/details?mode=charts) |
-| Nano | ONNX FP32 (AWS Graviton4 · 48-core) | ONNX | — | 51.13% | 29.16% | +0.02 | 109.43 | 136.39 | 110.6 | ~9 | [v-e8b](https://edgefirst.studio/public/validation/v-e8b/details?mode=charts) |
-| Nano | ONNX FP32 (AWS Graviton4 · 8-core) | ONNX | — | 51.13% | 29.16% | +0.02 | 329.09 | 352.64 | 23.4 | ~3 | [v-e72](https://edgefirst.studio/public/validation/v-e72/details?mode=charts) |
-| Nano | ONNX FP16 (Intel Core i9-13900F · 32-core) | ONNX | — | 50.27% | 28.69% | -0.45 | 104.40 | 207.52 | 17.5 | ~17 | [v-6dc](https://edgefirst.studio/public/validation/v-6dc/details?mode=charts) |
-| Nano | ONNX FP32 (Intel Core i9-13900F · 32-core) | ONNX | — | 51.13% | 29.14% | +0.00 | 55.84 | 83.99 | 47.7 | ~48 | [v-a49](https://edgefirst.studio/public/validation/v-a49/details?mode=charts) |
-| Nano | ONNX FP32 (Intel Xeon Platinum 8488C · 24-core) | ONNX | — | 51.13% | 29.14% | +0.00 | 102.37 | 144.94 | 98.4 | ~10 | [v-e90](https://edgefirst.studio/public/validation/v-e90/details?mode=charts) |
-| Nano | ONNX FP32 (Intel Xeon Platinum 8488C · 4-core) | ONNX | — | 51.13% | 29.14% | +0.00 | 262.65 | 316.10 | 27.5 | ~4 | [v-e7c](https://edgefirst.studio/public/validation/v-e7c/details?mode=charts) |
-| Nano | ONNX FP32 (CUDA) | ONNX | — | 51.13% | 29.14% | ref | 8.91 | 39.49 | 121.3 | ~133 | [v-e87](https://edgefirst.studio/public/validation/v-e87/details?mode=charts) |
-| Nano | ONNX FP32 (CUDA) | ONNX | — | 51.11% | 29.14% | +0.00 | 10.80 | 21.79 | 242.1 | ~242 | [v-a8b](https://edgefirst.studio/public/validation/v-a8b/details?mode=charts) |
-| Nano | ONNX FP16 (CUDA) | ONNX | — | 51.11% | 29.15% | +0.01 | 7.96 | 20.15 | 302.9 | ~303 | [v-aa0](https://edgefirst.studio/public/validation/v-aa0/details?mode=charts) |
-| Nano | Apple M2 Max — CoreML Neural Engine (FP16) | ONNX | — | 49.88% | 28.46% | -0.68 | 2.74 | 10.21 | 422.9 | ~424 | [v-72c](https://edgefirst.studio/public/validation/v-72c/details?mode=charts) |
-| Nano | Apple M2 Max — CoreML Metal GPU (FP16) | ONNX | — | 49.94% | 28.47% | -0.67 | 6.64 | 16.93 | 307.1 | ~307 | [v-736](https://edgefirst.studio/public/validation/v-736/details?mode=charts) |
-| Nano | Apple M2 Max — CoreML CPU (FP16) | ONNX | — | 50.78% | 28.92% | -0.22 | 21.74 | 29.90 | 87.2 | ~87 | [v-9f4](https://edgefirst.studio/public/validation/v-9f4/details?mode=charts) |
-| Nano | Apple iPhone 17 Pro — CoreML Neural Engine (FP16) | ONNX | — | 50.71% | 28.92% | -0.22 | 2.40 | 23.29 | 162.0 | ~63 | [v-f63](https://edgefirst.studio/public/validation/v-f63/details?mode=charts) |
-| Nano | Apple iPhone 17 Pro — CoreML Metal GPU (FP16) | ONNX | — | 50.76% | 28.92% | -0.22 | 8.86 | 30.07 | 135.3 | ~64 | [v-f62](https://edgefirst.studio/public/validation/v-f62/details?mode=charts) |
-| Nano | Apple iPhone 17 Pro — CoreML CPU (FP16) | ONNX | — | 50.78% | 28.95% | -0.19 | 23.23 | 50.04 | 72.5 | ~43 | [v-f60](https://edgefirst.studio/public/validation/v-f60/details?mode=charts) |
-| Nano | Apple iPhone 15 Pro — CoreML Neural Engine (FP16) | ONNX | — | 50.71% | 28.92% | -0.22 | 2.55 | 29.76 | 122.8 | ~49 | [v-f0d](https://edgefirst.studio/public/validation/v-f0d/details?mode=charts) |
-| Nano | Apple iPhone 15 Pro — CoreML Metal GPU (FP16) | ONNX | — | 50.76% | 28.92% | -0.22 | 16.81 | 52.22 | 79.3 | ~37 | [v-f0b](https://edgefirst.studio/public/validation/v-f0b/details?mode=charts) |
-| Nano | Apple iPhone 15 Pro — CoreML CPU (FP16) | ONNX | — | 50.79% | 28.96% | -0.18 | 28.84 | 68.76 | 56.3 | ~34 | [v-eff](https://edgefirst.studio/public/validation/v-eff/details?mode=charts) |
-| Nano | Samsung Galaxy S26 Ultra — Qualcomm Hexagon NPU (INT8) | ONNX · QNN EP | Smart | 48.59% | 27.45% | -1.69 | 2.14 | 40.25 | 92.9 | ~165 | [v-109c](https://edgefirst.studio/public/validation/v-109c/details?mode=charts) |
-| Nano | Samsung Galaxy S26 Ultra — Qualcomm Hexagon NPU (INT8) | ONNX · QNN EP | Logical | 37.92% | 18.50% | -10.64 | 2.06 | 13.32 | 347.1 | ~445 | [v-109d](https://edgefirst.studio/public/validation/v-109d/details?mode=charts) |
-| Nano | Samsung Galaxy S26 Ultra — Qualcomm Hexagon NPU (INT16) | ONNX · QNN EP | Smart | 50.24% | 28.60% | -0.54 | 4.99 | 52.18 | 82.3 | ~128 | [v-109a](https://edgefirst.studio/public/validation/v-109a/details?mode=charts) |
-| Nano | Samsung Galaxy S26 Ultra — Qualcomm Hexagon NPU (INT16) | ONNX · QNN EP | Logical | 49.37% | 10.81% | -18.33 | 5.41 | 29.82 | 135.1 | ~188 | [v-109b](https://edgefirst.studio/public/validation/v-109b/details?mode=charts) |
-| Nano | Samsung Galaxy S26 Ultra — CPU | LiteRT | Smart | 49.34% | 27.89% | -1.25 | 65.42 | 130.25 | 56.6 | ~15 | [v-fa9](https://edgefirst.studio/public/validation/v-fa9/details?mode=charts) |
-| Nano | NXP i.MX 8M Plus + VeriSilicon NPU (FRDM) | LiteRT | Combined | 37.94% | 21.74% | -7.40 | 82.14 | 123.21 | 10.7 | ~12 | [v-8d5](https://edgefirst.studio/public/validation/v-8d5/details?mode=charts) |
-| Nano | NXP i.MX 8M Plus + VeriSilicon NPU (FRDM) | LiteRT | Logical | 37.94% | 21.74% | -7.40 | 82.05 | 123.25 | 10.7 | ~12 | [v-8d6](https://edgefirst.studio/public/validation/v-8d6/details?mode=charts) |
-| Nano | NXP i.MX 8M Plus + VeriSilicon NPU (FRDM) | LiteRT | Smart | 48.83% | 27.59% | -1.55 | 79.61 | 199.44 | 10.6 | ~12 | [v-8d7](https://edgefirst.studio/public/validation/v-8d7/details?mode=charts) |
-| Nano | NXP i.MX 8M Plus + VeriSilicon NPU (Verdin) — latency | LiteRT | Logical | 37.94% | 21.74% | -7.40 | 83.28 | 135.65 | 10.5 | ~12 | [v-c84](https://edgefirst.studio/public/validation/v-c84/details?mode=charts) |
-| Nano | NXP i.MX 8M Plus + VeriSilicon NPU (Verdin) — throughput | LiteRT | Smart | 48.83% | 27.59% | -1.55 | 79.26 | 217.40 | 10.4 | ~12 | [v-c83](https://edgefirst.studio/public/validation/v-c83/details?mode=charts) |
-| Nano | NXP i.MX 95 + eIQ Neutron NPU — latency | LiteRT | Smart | 49.63% | 27.97% | -1.17 | 20.09 | 117.97 | 22.2 | ~14 | [v-e32](https://edgefirst.studio/public/validation/v-e32/details?mode=charts) |
-| Nano | NXP i.MX 95 + eIQ Neutron NPU — throughput | LiteRT | Smart | 49.62% | 27.97% | -1.17 | 24.58 | 186.87 | 23.6 | ~24 | [v-e33](https://edgefirst.studio/public/validation/v-e33/details?mode=charts) |
-| Nano | NXP i.MX 95 + eIQ Neutron NPU (FRDM) — latency | LiteRT | Smart | 48.87% | 27.63% | -1.51 | 20.62 | 116.40 | 20.8 | ~13 | [v-8cf](https://edgefirst.studio/public/validation/v-8cf/details?mode=charts) |
-| Nano | NXP i.MX 95 + eIQ Neutron NPU (FRDM) — latency | LiteRT | Logical | 38.09% | 21.84% | -7.30 | 49.14 | 73.69 | 19.4 | ~20 | [v-8cc](https://edgefirst.studio/public/validation/v-8cc/details?mode=charts) |
-| Nano | NXP i.MX 95 + eIQ Neutron NPU (FRDM) — latency | LiteRT | Combined | 38.09% | 21.84% | -7.30 | 48.82 | 73.23 | 19.5 | ~20 | [v-8ca](https://edgefirst.studio/public/validation/v-8ca/details?mode=charts) |
-| Nano | NXP i.MX 95 + eIQ Neutron NPU (FRDM) — throughput | LiteRT | Logical | 38.09% | 21.84% | -7.30 | 93.69 | 139.74 | 39.9 | ~40 | [v-8ce](https://edgefirst.studio/public/validation/v-8ce/details?mode=charts) |
-| Nano | NXP i.MX 95 + eIQ Neutron NPU (FRDM) — throughput | LiteRT | Combined | 38.09% | 21.84% | -7.30 | 93.37 | 139.14 | 40.3 | ~40 | [v-8cb](https://edgefirst.studio/public/validation/v-8cb/details?mode=charts) |
-| Nano | NXP i.MX 95 + eIQ Neutron NPU (Phytec) — throughput | LiteRT | Smart | 48.87% | 27.63% | -1.51 | 24.83 | 166.24 | 23.7 | ~24 | [v-8a8](https://edgefirst.studio/public/validation/v-8a8/details?mode=charts) |
-| Nano | NXP i.MX 95 + eIQ Neutron NPU (Verdin) — latency | LiteRT | Smart | 49.62% | 27.96% | -1.18 | 20.85 | 123.54 | 21.3 | ~14 | [v-e26](https://edgefirst.studio/public/validation/v-e26/details?mode=charts) |
-| Nano | NXP i.MX 95 + eIQ Neutron NPU (Verdin) — latency | LiteRT | Logical | 38.58% | 22.11% | -7.03 | 47.51 | 69.79 | 20.0 | ~20 | [v-bcf](https://edgefirst.studio/public/validation/v-bcf/details?mode=charts) |
-| Nano | NXP i.MX 95 + eIQ Neutron NPU (Verdin) — throughput | LiteRT | Smart | 49.63% | 27.97% | -1.17 | 25.98 | 199.50 | 22.6 | ~23 | [v-e0d](https://edgefirst.studio/public/validation/v-e0d/details?mode=charts) |
-| Nano | NXP Ara240 (FRDM) — latency | Ara DVM | Smart | 47.22% | 26.62% | -2.52 | 9.63 | 42.97 | 50.4 | ~56 | [v-a18](https://edgefirst.studio/public/validation/v-a18/details?mode=charts) |
-| Nano | NXP Ara240 (FRDM) — throughput | Ara DVM | Smart | 47.22% | 26.62% | -2.52 | 9.80 | 62.62 | 51.7 | ~226 | [v-a19](https://edgefirst.studio/public/validation/v-a19/details?mode=charts) |
-| Nano | Raspberry Pi 5 + Hailo-8L NPU | Hailo HEF | — | 49.28% | 28.02% | -1.12 | 16.18 | 38.25 | 53.7 | ~54 | [v-8e5](https://edgefirst.studio/public/validation/v-8e5/details?mode=charts) |
-| Nano | NVIDIA Jetson Orin Nano (TensorRT FP16) | TensorRT | — | 51.15% | 29.18% | +0.04 | 6.20 | 58.75 | 75.9 | ~60 | [v-91d](https://edgefirst.studio/public/validation/v-91d/details?mode=charts) |
-| Small | ONNX FP32 (AWS Graviton · 4-core) | ONNX | — | 60.07% | 34.82% | -0.01 | 1126.77 | 1155.09 | 3.5 | ~1 | [v-e5e](https://edgefirst.studio/public/validation/v-e5e/details?mode=charts) |
-| Small | ONNX FP32 (AWS Graviton4 · 48-core) | ONNX | — | 60.07% | 34.82% | -0.01 | 298.30 | 315.10 | 42.7 | ~3 | [v-e92](https://edgefirst.studio/public/validation/v-e92/details?mode=charts) |
-| Small | ONNX FP32 (AWS Graviton4 · 8-core) | ONNX | — | 60.07% | 34.82% | -0.01 | 918.98 | 933.58 | 8.6 | ~1 | [v-e8c](https://edgefirst.studio/public/validation/v-e8c/details?mode=charts) |
-| Small | ONNX FP32 (Intel Core i9-13900F · 32-core) | ONNX | — | 60.03% | 34.83% | +0.00 | 141.45 | 164.36 | 19.8 | ~20 | [v-a50](https://edgefirst.studio/public/validation/v-a50/details?mode=charts) |
-| Small | ONNX FP32 (Intel Xeon Platinum 8488C · 24-core) | ONNX | — | 60.03% | 34.83% | +0.00 | 253.09 | 289.72 | 48.8 | ~4 | [v-e94](https://edgefirst.studio/public/validation/v-e94/details?mode=charts) |
-| Small | ONNX FP32 (Intel Xeon Platinum 8488C · 4-core) | ONNX | — | 60.03% | 34.83% | +0.00 | 611.12 | 644.12 | 12.8 | ~2 | [v-e7e](https://edgefirst.studio/public/validation/v-e7e/details?mode=charts) |
-| Small | ONNX FP32 (CUDA) | ONNX | — | 60.02% | 34.83% | ref | 14.35 | 40.08 | 137.3 | ~148 | [v-e91](https://edgefirst.studio/public/validation/v-e91/details?mode=charts) |
-| Small | ONNX FP32 (CUDA) | ONNX | — | 60.02% | 34.83% | +0.00 | 20.42 | 31.47 | 151.7 | ~152 | [v-a92](https://edgefirst.studio/public/validation/v-a92/details?mode=charts) |
-| Small | ONNX FP16 (CUDA) | ONNX | — | 60.03% | 34.83% | +0.00 | 13.36 | 26.01 | 222.5 | ~222 | [v-aa7](https://edgefirst.studio/public/validation/v-aa7/details?mode=charts) |
-| Small | Apple M2 Max — CoreML Neural Engine (FP16) | ONNX | — | 59.62% | 34.65% | -0.18 | 6.18 | 13.83 | 261.8 | ~262 | [v-9e9](https://edgefirst.studio/public/validation/v-9e9/details?mode=charts) |
-| Small | Apple M2 Max — CoreML Metal GPU (FP16) | ONNX | — | 59.61% | 34.58% | -0.25 | 21.34 | 29.86 | 133.8 | ~134 | [v-9e8](https://edgefirst.studio/public/validation/v-9e8/details?mode=charts) |
-| Small | Apple M2 Max — CoreML CPU (FP16) | ONNX | — | 59.57% | 34.54% | -0.29 | 41.25 | 49.04 | 47.0 | ~47 | [v-9e7](https://edgefirst.studio/public/validation/v-9e7/details?mode=charts) |
-| Small | Apple iPhone 17 Pro — CoreML Neural Engine (FP16) | ONNX | — | 59.63% | 34.62% | -0.21 | 14.12 | 32.36 | 116.9 | ~71 | [v-f6d](https://edgefirst.studio/public/validation/v-f6d/details?mode=charts) |
-| Small | Apple iPhone 17 Pro — CoreML Metal GPU (FP16) | ONNX | — | 59.61% | 34.58% | -0.25 | 31.05 | 46.84 | 60.7 | ~32 | [v-f6c](https://edgefirst.studio/public/validation/v-f6c/details?mode=charts) |
-| Small | Apple iPhone 17 Pro — CoreML CPU (FP16) | ONNX | — | 59.56% | 34.55% | -0.28 | 72.03 | 88.45 | 27.2 | ~14 | [v-f6b](https://edgefirst.studio/public/validation/v-f6b/details?mode=charts) |
-| Small | Apple iPhone 15 Pro — CoreML Neural Engine (FP16) | ONNX | — | 59.63% | 34.63% | -0.20 | 12.65 | 44.68 | 102.0 | ~44 | [v-f31](https://edgefirst.studio/public/validation/v-f31/details?mode=charts) |
-| Small | Apple iPhone 15 Pro — CoreML Metal GPU (FP16) | ONNX | — | 59.62% | 34.59% | -0.24 | 75.06 | 88.35 | 26.0 | ~13 | [v-f34](https://edgefirst.studio/public/validation/v-f34/details?mode=charts) |
-| Small | Apple iPhone 15 Pro — CoreML CPU (FP16) | ONNX | — | 59.57% | 34.55% | -0.28 | 82.33 | 104.36 | 23.7 | ~12 | [v-f02](https://edgefirst.studio/public/validation/v-f02/details?mode=charts) |
-| Small | Samsung Galaxy S26 Ultra — Qualcomm Hexagon NPU (INT8) | ONNX · QNN EP | Smart | 58.59% | 33.71% | -1.12 | 2.68 | 34.62 | 117.2 | ~191 | [v-10a0](https://edgefirst.studio/public/validation/v-10a0/details?mode=charts) |
-| Small | Samsung Galaxy S26 Ultra — Qualcomm Hexagon NPU (INT8) | ONNX · QNN EP | Logical | 45.32% | 21.43% | -13.40 | 2.60 | 14.51 | 343.0 | ~420 | [v-10a1](https://edgefirst.studio/public/validation/v-10a1/details?mode=charts) |
-| Small | Samsung Galaxy S26 Ultra — Qualcomm Hexagon NPU (INT16) | ONNX · QNN EP | Smart | 59.48% | 34.40% | -0.43 | 5.71 | 45.70 | 93.6 | ~142 | [v-109e](https://edgefirst.studio/public/validation/v-109e/details?mode=charts) |
-| Small | Samsung Galaxy S26 Ultra — Qualcomm Hexagon NPU (INT16) | ONNX · QNN EP | Logical | 58.47% | 12.41% | -22.42 | 6.28 | 31.90 | 144.6 | ~187 | [v-109f](https://edgefirst.studio/public/validation/v-109f/details?mode=charts) |
-| Small | Samsung Galaxy S26 Ultra — CPU | LiteRT | Smart | 58.89% | 34.06% | -0.77 | 170.21 | 219.54 | 31.9 | ~6 | [v-fe7](https://edgefirst.studio/public/validation/v-fe7/details?mode=charts) |
-| Small | NXP i.MX 8M Plus + VeriSilicon NPU (FRDM) | LiteRT | Combined | 45.39% | 26.94% | -7.89 | 149.32 | 191.21 | 6.2 | ~7 | [v-96d](https://edgefirst.studio/public/validation/v-96d/details?mode=charts) |
-| Small | NXP i.MX 8M Plus + VeriSilicon NPU (FRDM) | LiteRT | Logical | 45.39% | 26.94% | -7.89 | 149.34 | 190.17 | 6.2 | ~7 | [v-970](https://edgefirst.studio/public/validation/v-970/details?mode=charts) |
-| Small | NXP i.MX 8M Plus + VeriSilicon NPU (FRDM) | LiteRT | Smart | 58.45% | 33.86% | -0.97 | 147.05 | 226.42 | 6.3 | ~7 | [v-972](https://edgefirst.studio/public/validation/v-972/details?mode=charts) |
-| Small | NXP i.MX 8M Plus + VeriSilicon NPU (Verdin) — latency | LiteRT | Logical | 45.39% | 26.94% | -7.89 | 151.70 | 203.35 | 6.1 | ~7 | [v-c93](https://edgefirst.studio/public/validation/v-c93/details?mode=charts) |
-| Small | NXP i.MX 8M Plus + VeriSilicon NPU (Verdin) — throughput | LiteRT | Smart | 58.45% | 33.86% | -0.97 | 146.93 | 239.32 | 6.3 | ~7 | [v-c92](https://edgefirst.studio/public/validation/v-c92/details?mode=charts) |
-| Small | NXP i.MX 95 + eIQ Neutron NPU — latency | LiteRT | Smart | 59.07% | 34.06% | -0.77 | 49.66 | 101.69 | 19.2 | ~19 | [v-e36](https://edgefirst.studio/public/validation/v-e36/details?mode=charts) |
-| Small | NXP i.MX 95 + eIQ Neutron NPU — throughput | LiteRT | Smart | 59.07% | 34.06% | -0.77 | 377.59 | 436.63 | 20.9 | ~21 | [v-e37](https://edgefirst.studio/public/validation/v-e37/details?mode=charts) |
-| Small | NXP i.MX 95 + eIQ Neutron NPU (FRDM) — latency | LiteRT | Smart | 58.42% | 33.76% | -1.07 | 51.16 | 112.09 | 18.3 | ~19 | [v-941](https://edgefirst.studio/public/validation/v-941/details?mode=charts) |
-| Small | NXP i.MX 95 + eIQ Neutron NPU (FRDM) — latency | LiteRT | Logical | 45.37% | 26.80% | -8.03 | 87.78 | 112.37 | 11.0 | ~11 | [v-93e](https://edgefirst.studio/public/validation/v-93e/details?mode=charts) |
-| Small | NXP i.MX 95 + eIQ Neutron NPU (FRDM) — latency | LiteRT | Combined | 45.37% | 26.80% | -8.03 | 86.87 | 111.07 | 11.2 | ~11 | [v-93b](https://edgefirst.studio/public/validation/v-93b/details?mode=charts) |
-| Small | NXP i.MX 95 + eIQ Neutron NPU (FRDM) — throughput | LiteRT | Smart | 58.42% | 33.76% | -1.07 | 194.35 | 262.79 | 19.8 | ~20 | [v-943](https://edgefirst.studio/public/validation/v-943/details?mode=charts) |
-| Small | NXP i.MX 95 + eIQ Neutron NPU (FRDM) — throughput | LiteRT | Logical | 45.37% | 26.80% | -8.03 | 190.56 | 214.93 | 20.7 | ~21 | [v-940](https://edgefirst.studio/public/validation/v-940/details?mode=charts) |
-| Small | NXP i.MX 95 + eIQ Neutron NPU (FRDM) — throughput | LiteRT | Combined | 45.37% | 26.80% | -8.03 | 190.30 | 214.39 | 20.7 | ~21 | [v-93d](https://edgefirst.studio/public/validation/v-93d/details?mode=charts) |
-| Small | NXP i.MX 95 + eIQ Neutron NPU (Phytec) — throughput | LiteRT | Smart | 58.42% | 33.76% | -1.07 | 185.95 | 245.69 | 20.8 | ~21 | [v-8fd](https://edgefirst.studio/public/validation/v-8fd/details?mode=charts) |
-| Small | NXP i.MX 95 + eIQ Neutron NPU (Verdin) — latency | LiteRT | Smart | 59.07% | 34.06% | -0.77 | 50.53 | 105.70 | 18.8 | ~19 | [v-e27](https://edgefirst.studio/public/validation/v-e27/details?mode=charts) |
-| Small | NXP i.MX 95 + eIQ Neutron NPU (Verdin) — latency | LiteRT | Logical | 45.84% | 27.17% | -7.66 | 77.67 | 99.84 | 12.5 | ~13 | [v-bdd](https://edgefirst.studio/public/validation/v-bdd/details?mode=charts) |
-| Small | NXP i.MX 95 + eIQ Neutron NPU (Verdin) — throughput | LiteRT | Smart | 59.07% | 34.06% | -0.77 | 388.03 | 450.48 | 20.3 | ~20 | [v-e14](https://edgefirst.studio/public/validation/v-e14/details?mode=charts) |
-| Small | NXP Ara240 (FRDM) — latency | Ara DVM | Smart | 56.31% | 32.62% | -2.21 | 15.64 | 42.24 | 53.6 | ~54 | [v-a26](https://edgefirst.studio/public/validation/v-a26/details?mode=charts) |
-| Small | NXP Ara240 (FRDM) — throughput | Ara DVM | Smart | 56.32% | 32.63% | -2.20 | 16.15 | 66.47 | 64.7 | ~94 | [v-a27](https://edgefirst.studio/public/validation/v-a27/details?mode=charts) |
-| Small | Raspberry Pi 5 + Hailo-8L NPU | Hailo HEF | — | 58.40% | 33.68% | -1.15 | 42.29 | 61.72 | 22.7 | ~23 | [v-8ef](https://edgefirst.studio/public/validation/v-8ef/details?mode=charts) |
-| Small | NVIDIA Jetson Orin Nano (TensorRT FP16) | TensorRT | — | 60.05% | 34.84% | +0.01 | 14.01 | 62.73 | 86.6 | ~68 | [v-925](https://edgefirst.studio/public/validation/v-925/details?mode=charts) |
-| Medium | ONNX FP32 (AWS Graviton · 4-core) | ONNX | — | 65.07% | 38.13% | -0.01 | 2806.49 | 2832.98 | 1.4 | ~0 | [v-e5f](https://edgefirst.studio/public/validation/v-e5f/details?mode=charts) |
-| Medium | ONNX FP32 (AWS Graviton4 · 48-core) | ONNX | — | 65.07% | 38.13% | -0.01 | 724.39 | 740.88 | 17.6 | ~1 | [v-e9e](https://edgefirst.studio/public/validation/v-e9e/details?mode=charts) |
-| Medium | ONNX FP32 (AWS Graviton4 · 8-core) | ONNX | — | 65.07% | 38.13% | -0.01 | 2244.67 | 2256.58 | 3.6 | ~0 | [v-e76](https://edgefirst.studio/public/validation/v-e76/details?mode=charts) |
-| Medium | ONNX FP32 (Intel Core i9-13900F · 32-core) | ONNX | — | 65.08% | 38.14% | +0.00 | 324.10 | 345.48 | 8.9 | ~9 | [v-a57](https://edgefirst.studio/public/validation/v-a57/details?mode=charts) |
-| Medium | ONNX FP32 (Intel Xeon Platinum 8488C · 24-core) | ONNX | — | 65.08% | 38.14% | +0.00 | 588.29 | 623.24 | 21.2 | ~2 | [v-e99](https://edgefirst.studio/public/validation/v-e99/details?mode=charts) |
-| Medium | ONNX FP32 (Intel Xeon Platinum 8488C · 4-core) | ONNX | — | 65.08% | 38.14% | +0.00 | 1421.89 | 1448.31 | 5.6 | ~1 | [v-e81](https://edgefirst.studio/public/validation/v-e81/details?mode=charts) |
-| Medium | ONNX FP32 (CUDA) | ONNX | — | 65.08% | 38.14% | ref | 29.42 | 48.85 | 112.7 | ~113 | [v-e8f](https://edgefirst.studio/public/validation/v-e8f/details?mode=charts) |
-| Medium | ONNX FP32 (CUDA) | ONNX | — | 65.08% | 38.15% | +0.01 | 55.06 | 65.30 | 62.1 | ~62 | [v-a99](https://edgefirst.studio/public/validation/v-a99/details?mode=charts) |
-| Medium | ONNX FP16 (CUDA) | ONNX | — | 65.06% | 38.15% | +0.01 | 28.47 | 41.89 | 118.2 | ~118 | [v-aae](https://edgefirst.studio/public/validation/v-aae/details?mode=charts) |
-| Medium | Apple M2 Max — CoreML Neural Engine (FP16) | ONNX | — | 63.08% | 37.10% | -1.04 | 17.13 | 23.51 | 111.3 | ~111 | [v-72e](https://edgefirst.studio/public/validation/v-72e/details?mode=charts) |
-| Medium | Apple M2 Max — CoreML Metal GPU (FP16) | ONNX | — | 64.09% | 37.54% | -0.60 | 50.64 | 58.53 | 58.0 | ~58 | [v-9df](https://edgefirst.studio/public/validation/v-9df/details?mode=charts) |
-| Medium | Apple M2 Max — CoreML CPU (FP16) | ONNX | — | 64.07% | 37.51% | -0.63 | 78.49 | 86.82 | 25.1 | ~25 | [v-9de](https://edgefirst.studio/public/validation/v-9de/details?mode=charts) |
-| Medium | Apple iPhone 17 Pro — CoreML Neural Engine (FP16) | ONNX | — | 64.05% | 37.62% | -0.52 | 25.96 | 36.33 | 73.9 | ~39 | [v-f74](https://edgefirst.studio/public/validation/v-f74/details?mode=charts) |
-| Medium | Apple iPhone 17 Pro — CoreML Metal GPU (FP16) | ONNX | — | 64.09% | 37.55% | -0.59 | 67.60 | 77.83 | 29.0 | ~15 | [v-f76](https://edgefirst.studio/public/validation/v-f76/details?mode=charts) |
-| Medium | Apple iPhone 17 Pro — CoreML CPU (FP16) | ONNX | — | 64.06% | 37.51% | -0.63 | 163.92 | 179.34 | 12.1 | ~6 | [v-f75](https://edgefirst.studio/public/validation/v-f75/details?mode=charts) |
-| Medium | Apple iPhone 15 Pro — CoreML Neural Engine (FP16) | ONNX | — | 64.05% | 37.62% | -0.52 | 32.97 | 48.55 | 57.9 | ~30 | [v-f43](https://edgefirst.studio/public/validation/v-f43/details?mode=charts) |
-| Medium | Apple iPhone 15 Pro — CoreML Metal GPU (FP16) | ONNX | — | 64.10% | 37.53% | -0.61 | 181.18 | 199.20 | 10.8 | ~6 | [v-f4e](https://edgefirst.studio/public/validation/v-f4e/details?mode=charts) |
-| Medium | Apple iPhone 15 Pro — CoreML CPU (FP16) | ONNX | — | 64.06% | 37.53% | -0.61 | 181.67 | 204.14 | 10.9 | ~6 | [v-f05](https://edgefirst.studio/public/validation/v-f05/details?mode=charts) |
-| Medium | Samsung Galaxy S26 Ultra — Qualcomm Hexagon NPU (INT8) | ONNX · QNN EP | Smart | 63.95% | 37.29% | -0.85 | 5.94 | 40.46 | 109.8 | ~174 | [v-10a4](https://edgefirst.studio/public/validation/v-10a4/details?mode=charts) |
-| Medium | Samsung Galaxy S26 Ultra — Qualcomm Hexagon NPU (INT8) | ONNX · QNN EP | Logical | 47.02% | 24.48% | -13.66 | 5.65 | 18.87 | 258.3 | ~305 | [v-10a5](https://edgefirst.studio/public/validation/v-10a5/details?mode=charts) |
-| Medium | Samsung Galaxy S26 Ultra — Qualcomm Hexagon NPU (INT16) | ONNX · QNN EP | Smart | 64.77% | 37.82% | -0.32 | 9.83 | 52.86 | 80.8 | ~114 | [v-10a2](https://edgefirst.studio/public/validation/v-10a2/details?mode=charts) |
-| Medium | Samsung Galaxy S26 Ultra — Qualcomm Hexagon NPU (INT16) | ONNX · QNN EP | Logical | 63.71% | 13.22% | -24.92 | 14.06 | 40.09 | 108.8 | ~125 | [v-10a3](https://edgefirst.studio/public/validation/v-10a3/details?mode=charts) |
-| Medium | Samsung Galaxy S26 Ultra — CPU | LiteRT | Smart | 63.62% | 37.20% | -0.94 | 423.67 | 467.09 | 13.7 | ~2 | [v-101f](https://edgefirst.studio/public/validation/v-101f/details?mode=charts) |
-| Medium | NXP i.MX 8M Plus + VeriSilicon NPU (FRDM) | LiteRT | Combined | 46.05% | 27.68% | -10.46 | 272.63 | 316.01 | 3.5 | ~4 | [v-9a6](https://edgefirst.studio/public/validation/v-9a6/details?mode=charts) |
-| Medium | NXP i.MX 8M Plus + VeriSilicon NPU (FRDM) | LiteRT | Logical | 46.05% | 27.68% | -10.46 | 272.73 | 313.04 | 3.5 | ~4 | [v-9a7](https://edgefirst.studio/public/validation/v-9a7/details?mode=charts) |
-| Medium | NXP i.MX 8M Plus + VeriSilicon NPU (FRDM) | LiteRT | Smart | 62.77% | 36.63% | -1.51 | 267.65 | 343.19 | 3.6 | ~4 | [v-9a8](https://edgefirst.studio/public/validation/v-9a8/details?mode=charts) |
-| Medium | NXP i.MX 8M Plus + VeriSilicon NPU (Verdin) — latency | LiteRT | Logical | 46.05% | 27.68% | -10.46 | 276.12 | 327.36 | 3.5 | ~4 | [v-ca7](https://edgefirst.studio/public/validation/v-ca7/details?mode=charts) |
-| Medium | NXP i.MX 8M Plus + VeriSilicon NPU (Verdin) — throughput | LiteRT | Smart | 62.77% | 36.63% | -1.51 | 270.51 | 363.03 | 3.6 | ~4 | [v-ca6](https://edgefirst.studio/public/validation/v-ca6/details?mode=charts) |
-| Medium | NXP i.MX 95 + eIQ Neutron NPU — latency | LiteRT | Smart | 64.23% | 37.49% | -0.65 | 125.91 | 186.36 | 7.8 | ~8 | [v-e43](https://edgefirst.studio/public/validation/v-e43/details?mode=charts) |
-| Medium | NXP i.MX 95 + eIQ Neutron NPU — throughput | LiteRT | Smart | 64.22% | 37.49% | -0.65 | 986.97 | 1045.73 | 8.1 | ~8 | [v-e44](https://edgefirst.studio/public/validation/v-e44/details?mode=charts) |
-| Medium | NXP i.MX 95 + eIQ Neutron NPU (FRDM) — latency | LiteRT | Smart | 62.66% | 36.61% | -1.53 | 127.25 | 177.98 | 7.6 | ~8 | [v-98f](https://edgefirst.studio/public/validation/v-98f/details?mode=charts) |
-| Medium | NXP i.MX 95 + eIQ Neutron NPU (FRDM) — latency | LiteRT | Logical | 46.14% | 27.71% | -10.43 | 162.78 | 185.83 | 6.0 | ~6 | [v-98c](https://edgefirst.studio/public/validation/v-98c/details?mode=charts) |
-| Medium | NXP i.MX 95 + eIQ Neutron NPU (FRDM) — latency | LiteRT | Combined | 46.14% | 27.71% | -10.43 | 162.62 | 185.52 | 6.0 | ~6 | [v-98a](https://edgefirst.studio/public/validation/v-98a/details?mode=charts) |
-| Medium | NXP i.MX 95 + eIQ Neutron NPU (FRDM) — throughput | LiteRT | Smart | 62.66% | 36.61% | -1.53 | 495.23 | 548.82 | 8.0 | ~8 | [v-991](https://edgefirst.studio/public/validation/v-991/details?mode=charts) |
-| Medium | NXP i.MX 95 + eIQ Neutron NPU (FRDM) — throughput | LiteRT | Logical | 46.14% | 27.71% | -10.43 | 490.67 | 516.43 | 8.1 | ~8 | [v-98e](https://edgefirst.studio/public/validation/v-98e/details?mode=charts) |
-| Medium | NXP i.MX 95 + eIQ Neutron NPU (FRDM) — throughput | LiteRT | Combined | 46.14% | 27.71% | -10.43 | 490.30 | 515.70 | 8.1 | ~8 | [v-98b](https://edgefirst.studio/public/validation/v-98b/details?mode=charts) |
-| Medium | NXP i.MX 95 + eIQ Neutron NPU (Phytec) — throughput | LiteRT | Smart | 62.66% | 36.61% | -1.53 | 484.09 | 531.70 | 8.2 | ~8 | [v-959](https://edgefirst.studio/public/validation/v-959/details?mode=charts) |
-| Medium | NXP i.MX 95 + eIQ Neutron NPU (Verdin) — latency | LiteRT | Smart | 64.23% | 37.49% | -0.65 | 125.61 | 188.01 | 7.8 | ~8 | [v-e28](https://edgefirst.studio/public/validation/v-e28/details?mode=charts) |
-| Medium | NXP i.MX 95 + eIQ Neutron NPU (Verdin) — latency | LiteRT | Logical | 47.18% | 28.34% | -9.80 | 152.43 | 174.20 | 6.5 | ~7 | [v-beb](https://edgefirst.studio/public/validation/v-beb/details?mode=charts) |
-| Medium | NXP i.MX 95 + eIQ Neutron NPU (Verdin) — throughput | LiteRT | Smart | 64.22% | 37.49% | -0.65 | 983.05 | 1044.65 | 8.1 | ~8 | [v-e1b](https://edgefirst.studio/public/validation/v-e1b/details?mode=charts) |
-| Medium | NXP Ara240 (FRDM) — latency | Ara DVM | Smart | 58.60% | 34.00% | -4.14 | 31.02 | 53.88 | 30.9 | ~31 | [v-a34](https://edgefirst.studio/public/validation/v-a34/details?mode=charts) |
-| Medium | NXP Ara240 (FRDM) — throughput | Ara DVM | Smart | 58.59% | 33.99% | -4.15 | 31.15 | 55.73 | 38.1 | ~38 | [v-a35](https://edgefirst.studio/public/validation/v-a35/details?mode=charts) |
-| Medium | Raspberry Pi 5 + Hailo-8L NPU | Hailo HEF | — | 63.27% | 37.03% | -1.11 | 66.60 | 86.17 | 13.3 | ~13 | [v-931](https://edgefirst.studio/public/validation/v-931/details?mode=charts) |
-| Medium | NVIDIA Jetson Orin Nano (TensorRT FP16) | TensorRT | — | 65.05% | 38.14% | +0.00 | 69.82 | 99.66 | 56.7 | ~57 | [v-92e](https://edgefirst.studio/public/validation/v-92e/details?mode=charts) |
+- **Platform-label suffixes.** `(EVK)` / `(FRDM)` / `(Phytec)` / `(Verdin)` / `(Ezurio)` name the NXP i.MX development board a session ran on. `— latency` / `— throughput` mark the two pipeline configurations the NXP i.MX 95 Neutron and NXP Ara240 targets run: the **latency** pipeline runs inference serially for the lowest per-frame latency; the **throughput** pipeline runs multiple inference workers for the highest FPS, which raises per-frame latency in exchange. Rows with neither suffix run a single pipeline.
+- **End-to-end (ms)** is the per-image latency from the moment a frame is available to the moment its results are ready: preprocess → inference → postprocess. Image acquisition (camera capture, or file load and JPEG decode), mask encoding and writing results to disk are excluded.
+- **Realized FPS** is the *measured* steady-state throughput — the rate at which final results are actually delivered over the full validation pipeline. It normally exceeds `1000 / end-to-end` because the runtime overlaps stages across frames. Validation runs decode a JPEG per image and evaluate at a `0.001` confidence threshold to capture every detection for mAP; a live camera pipeline at a deployment threshold of `0.25–0.75` carries a lighter host and postprocess load.
+
+| Size | Platform | Runtime | Decode | Box mAP@0.5 | Mask mAP@0.5-0.95 | Δ mask vs FP32 (pp) | End-to-end (ms) | Realized FPS | Source |
+|------|----------|---------|--------|-------------|-------------------|---------------------|-----------------|--------------|--------|
+| Nano | ONNX FP32 (AWS Graviton · 4-core) | ONNX | — | 51.13% | 29.15% | +0.01 | 422.02 | 10.0 | [v-e5d](https://edgefirst.studio/public/validation/v-e5d/details?mode=charts) · 2026 Q2 |
+| Nano | ONNX FP16 (AWS Graviton) | ONNX | — | 50.29% | 28.67% | -0.47 | 357.00 | 6.9 | [v-711](https://edgefirst.studio/public/validation/v-711/details?mode=charts) · 2026 Q2 |
+| Nano | ONNX FP32 (AWS Graviton4 · 48-core) | ONNX | — | 51.13% | 29.16% | +0.02 | 133.25 | 110.6 | [v-e8b](https://edgefirst.studio/public/validation/v-e8b/details?mode=charts) · 2026 Q2 |
+| Nano | ONNX FP32 (AWS Graviton4 · 8-core) | ONNX | — | 51.13% | 29.16% | +0.02 | 349.48 | 23.4 | [v-e72](https://edgefirst.studio/public/validation/v-e72/details?mode=charts) · 2026 Q2 |
+| Nano | ONNX FP16 (Intel Core i9-13900F · 32-core) | ONNX | — | 50.27% | 28.69% | -0.45 | 200.54 | 17.5 | [v-6dc](https://edgefirst.studio/public/validation/v-6dc/details?mode=charts) · 2026 Q2 |
+| Nano | ONNX FP32 (Intel Core i9-13900F · 32-core) | ONNX | — | 51.13% | 29.14% | +0.00 | 78.75 | 47.7 | [v-a49](https://edgefirst.studio/public/validation/v-a49/details?mode=charts) · 2026 Q2 |
+| Nano | ONNX FP32 (Intel Xeon Platinum 8488C · 24-core) | ONNX | — | 51.13% | 29.14% | +0.00 | 140.27 | 98.4 | [v-e90](https://edgefirst.studio/public/validation/v-e90/details?mode=charts) · 2026 Q2 |
+| Nano | ONNX FP32 (Intel Xeon Platinum 8488C · 4-core) | ONNX | — | 51.13% | 29.14% | +0.00 | 310.18 | 27.5 | [v-e7c](https://edgefirst.studio/public/validation/v-e7c/details?mode=charts) · 2026 Q2 |
+| Nano | ONNX FP32 (CUDA · A10G) | ONNX | — | 51.13% | 29.14% | ref | 34.66 | 121.3 | [v-e87](https://edgefirst.studio/public/validation/v-e87/details?mode=charts) · 2026 Q2 |
+| Nano | ONNX FP32 (CUDA · RTX 4060) | ONNX | — | 51.11% | 29.14% | +0.00 | 18.38 | 242.1 | [v-a8b](https://edgefirst.studio/public/validation/v-a8b/details?mode=charts) · 2026 Q2 |
+| Nano | ONNX FP16 (CUDA · RTX 4060) | ONNX | — | 51.11% | 29.15% | +0.01 | 16.66 | 302.9 | [v-aa0](https://edgefirst.studio/public/validation/v-aa0/details?mode=charts) · 2026 Q2 |
+| Nano | ONNX FP32 (CUDA · RTX 3090) | ONNX | — | 51.43% | 29.24% | ref | 26.38 | 194.8 | [v-1945](https://edgefirst.studio/public/validation/v-1945/details?mode=charts) |
+| Nano | ONNX FP16 (CUDA · RTX 3090) | ONNX | — | 51.41% | 29.23% | -0.01 | 21.74 | 239.7 | [v-19af](https://edgefirst.studio/public/validation/v-19af/details?mode=charts) |
+| Nano | Apple M2 Max — CoreML Neural Engine (FP16) | ONNX | — | 49.88% | 28.46% | -0.68 | 7.94 | 422.9 | [v-72c](https://edgefirst.studio/public/validation/v-72c/details?mode=charts) · 2026 Q2 |
+| Nano | Apple M2 Max — CoreML Metal GPU (FP16) | ONNX | — | 49.94% | 28.47% | -0.67 | 13.64 | 307.1 | [v-736](https://edgefirst.studio/public/validation/v-736/details?mode=charts) · 2026 Q2 |
+| Nano | Apple M2 Max — CoreML CPU (FP16) | ONNX | — | 50.78% | 28.92% | -0.22 | 27.43 | 87.2 | [v-9f4](https://edgefirst.studio/public/validation/v-9f4/details?mode=charts) · 2026 Q2 |
+| Nano | Apple iPhone 17 Pro — CoreML Neural Engine (FP16) | ONNX | — | 50.71% | 28.92% | -0.22 | 19.73 | 162.0 | [v-f63](https://edgefirst.studio/public/validation/v-f63/details?mode=charts) · 2026 Q2 |
+| Nano | Apple iPhone 17 Pro — CoreML Metal GPU (FP16) | ONNX | — | 50.76% | 28.92% | -0.22 | 26.00 | 135.3 | [v-f62](https://edgefirst.studio/public/validation/v-f62/details?mode=charts) · 2026 Q2 |
+| Nano | Apple iPhone 17 Pro — CoreML CPU (FP16) | ONNX | — | 50.78% | 28.95% | -0.19 | 44.08 | 72.5 | [v-f60](https://edgefirst.studio/public/validation/v-f60/details?mode=charts) · 2026 Q2 |
+| Nano | Apple iPhone 15 Pro — CoreML Neural Engine (FP16) | ONNX | — | 50.71% | 28.92% | -0.22 | 25.00 | 122.8 | [v-f0d](https://edgefirst.studio/public/validation/v-f0d/details?mode=charts) · 2026 Q2 |
+| Nano | Apple iPhone 15 Pro — CoreML Metal GPU (FP16) | ONNX | — | 50.76% | 28.92% | -0.22 | 45.63 | 79.3 | [v-f0b](https://edgefirst.studio/public/validation/v-f0b/details?mode=charts) · 2026 Q2 |
+| Nano | Apple iPhone 15 Pro — CoreML CPU (FP16) | ONNX | — | 50.79% | 28.96% | -0.18 | 60.68 | 56.3 | [v-eff](https://edgefirst.studio/public/validation/v-eff/details?mode=charts) · 2026 Q2 |
+| Nano | Samsung Galaxy S26 Ultra — Qualcomm Hexagon NPU (INT8) | ONNX · QNN EP | Smart | 48.59% | 27.45% | -1.69 | 36.18 | 92.9 | [v-109c](https://edgefirst.studio/public/validation/v-109c/details?mode=charts) · 2026 Q2 |
+| Nano | Samsung Galaxy S26 Ultra — Qualcomm Hexagon NPU (INT8) | ONNX · QNN EP | Logical | 37.92% | 18.50% | -10.64 | 10.32 | 347.1 | [v-109d](https://edgefirst.studio/public/validation/v-109d/details?mode=charts) · 2026 Q2 |
+| Nano | Samsung Galaxy S26 Ultra — Qualcomm Hexagon NPU (INT16) | ONNX · QNN EP | Smart | 50.24% | 28.60% | -0.54 | 46.67 | 82.3 | [v-109a](https://edgefirst.studio/public/validation/v-109a/details?mode=charts) · 2026 Q2 |
+| Nano | Samsung Galaxy S26 Ultra — Qualcomm Hexagon NPU (INT16) | ONNX · QNN EP | Logical | 49.37% | 10.81% | -18.33 | 25.15 | 135.1 | [v-109b](https://edgefirst.studio/public/validation/v-109b/details?mode=charts) · 2026 Q2 |
+| Nano | Samsung Galaxy S26 Ultra — CPU | LiteRT | Smart | 49.34% | 27.89% | -1.25 | 119.50 | 56.6 | [v-fa9](https://edgefirst.studio/public/validation/v-fa9/details?mode=charts) · 2026 Q2 |
+| Nano | NXP i.MX 8M Plus + VeriSilicon NPU (Ezurio) — latency | LiteRT | Logical | 48.18% | 27.35% | -1.89 | 328.11 | 4.3 | [v-18a9](https://edgefirst.studio/public/validation/v-18a9/details?mode=charts) |
+| Nano | NXP i.MX 8M Plus + VeriSilicon NPU (FRDM) | LiteRT | Combined | 39.30% | 22.47% | -6.77 | 102.25 | 10.8 | [v-18a1](https://edgefirst.studio/public/validation/v-18a1/details?mode=charts) |
+| Nano | NXP i.MX 8M Plus + VeriSilicon NPU (FRDM) | LiteRT | Logical | 48.18% | 27.35% | -1.89 | 110.62 | 10.6 | [v-18aa](https://edgefirst.studio/public/validation/v-18aa/details?mode=charts) |
+| Nano | NXP i.MX 8M Plus + VeriSilicon NPU (FRDM) | LiteRT | Smart | 49.82% | 28.01% | -1.23 | 156.27 | 11.1 | [v-18bc](https://edgefirst.studio/public/validation/v-18bc/details?mode=charts) |
+| Nano | NXP i.MX 8M Plus + VeriSilicon NPU (Phytec) — latency | LiteRT | Logical | 48.18% | 27.31% | -1.93 | 116.32 | 11.4 | [v-18ab](https://edgefirst.studio/public/validation/v-18ab/details?mode=charts) |
+| Nano | NXP i.MX 8M Plus + VeriSilicon NPU (Verdin) — latency | LiteRT | Logical | 48.18% | 27.35% | -1.89 | 115.72 | 10.3 | [v-18ac](https://edgefirst.studio/public/validation/v-18ac/details?mode=charts) |
+| Nano | NXP i.MX 8M Plus + VeriSilicon NPU (Verdin) — throughput | LiteRT | Smart | 48.83% | 27.59% | -1.55 | 184.42 | 10.4 | [v-c83](https://edgefirst.studio/public/validation/v-c83/details?mode=charts) · 2026 Q2 |
+| Nano | NXP i.MX 95 + eIQ Neutron NPU (Ezurio) — latency | LiteRT | Smart | 49.60% | 27.99% | -1.25 | 101.04 | 24.8 | [v-1628](https://edgefirst.studio/public/validation/v-1628/details?mode=charts) |
+| Nano | NXP i.MX 95 + eIQ Neutron NPU (Ezurio) — throughput | LiteRT | Smart | 49.60% | 27.99% | -1.25 | 159.61 | 26.6 | [v-18ff](https://edgefirst.studio/public/validation/v-18ff/details?mode=charts) |
+| Nano | NXP i.MX 95 + eIQ Neutron NPU (EVK) — latency | LiteRT | Smart | 49.72% | 28.10% | -1.14 | 89.10 | 24.9 | [v-180c](https://edgefirst.studio/public/validation/v-180c/details?mode=charts) |
+| Nano | NXP i.MX 95 + eIQ Neutron NPU (EVK) — latency | LiteRT | Logical | 48.20% | 27.47% | -1.77 | 62.28 | 21.6 | [v-1806](https://edgefirst.studio/public/validation/v-1806/details?mode=charts) |
+| Nano | NXP i.MX 95 + eIQ Neutron NPU (EVK) — latency | LiteRT | Combined | 39.19% | 22.52% | -6.72 | 60.11 | 20.5 | [v-1800](https://edgefirst.studio/public/validation/v-1800/details?mode=charts) |
+| Nano | NXP i.MX 95 + eIQ Neutron NPU (EVK) — throughput | LiteRT | Smart | 49.72% | 28.10% | -1.14 | 141.92 | 26.5 | [v-1812](https://edgefirst.studio/public/validation/v-1812/details?mode=charts) |
+| Nano | NXP i.MX 95 + eIQ Neutron NPU (EVK) — throughput | LiteRT | Logical | 48.20% | 27.47% | -1.77 | 135.25 | 41.1 | [v-180a](https://edgefirst.studio/public/validation/v-180a/details?mode=charts) |
+| Nano | NXP i.MX 95 + eIQ Neutron NPU (EVK) — throughput | LiteRT | Combined | 39.19% | 22.52% | -6.72 | 103.46 | 51.5 | [v-1804](https://edgefirst.studio/public/validation/v-1804/details?mode=charts) |
+| Nano | NXP i.MX 95 + eIQ Neutron NPU (Phytec) — throughput | LiteRT | Smart | 49.72% | 28.10% | -1.14 | 140.99 | 27.0 | [v-1775](https://edgefirst.studio/public/validation/v-1775/details?mode=charts) |
+| Nano | NXP i.MX 95 + eIQ Neutron NPU (Verdin) — latency | LiteRT | Smart | 49.60% | 27.99% | -1.25 | 105.39 | 23.9 | [v-162a](https://edgefirst.studio/public/validation/v-162a/details?mode=charts) |
+| Nano | NXP i.MX 95 + eIQ Neutron NPU (Verdin) — latency | LiteRT | Logical | 38.58% | 22.11% | -7.03 | 63.63 | 20.0 | [v-bcf](https://edgefirst.studio/public/validation/v-bcf/details?mode=charts) · 2026 Q2 |
+| Nano | NXP i.MX 95 + eIQ Neutron NPU (Verdin) — throughput | LiteRT | Smart | 49.60% | 27.99% | -1.25 | 166.29 | 25.3 | [v-1908](https://edgefirst.studio/public/validation/v-1908/details?mode=charts) |
+| Nano | NXP Ara240 (FRDM) — latency | Ara DVM | Smart | 48.81% | 27.69% | -1.55 | 49.40 | 43.4 | [v-160f](https://edgefirst.studio/public/validation/v-160f/details?mode=charts) |
+| Nano | NXP Ara240 (FRDM) — throughput | Ara DVM | Smart | 48.81% | 27.69% | -1.55 | 62.93 | 46.6 | [v-18df](https://edgefirst.studio/public/validation/v-18df/details?mode=charts) |
+| Nano | Raspberry Pi 5 + Hailo-8L NPU | Hailo HEF | — | 49.54% | 28.11% | -1.13 | 29.94 | 52.8 | [v-153f](https://edgefirst.studio/public/validation/v-153f/details?mode=charts) |
+| Nano | NVIDIA Jetson Orin Nano (TensorRT FP16) | TensorRT | — | 51.39% | 29.27% | +0.03 | 73.82 | 87.2 | [v-1759](https://edgefirst.studio/public/validation/v-1759/details?mode=charts) |
+| Nano | NVIDIA Jetson Orin Nano (TensorRT INT8) | TensorRT | — | 50.10% | 28.47% | -0.77 | 72.69 | 83.7 | [v-175b](https://edgefirst.studio/public/validation/v-175b/details?mode=charts) |
+| Nano | Qualcomm Dragonwing IQ-9075 — Hexagon NPU (INT8) | ONNX · QNN EP | Smart | 50.58% | 28.70% | -0.54 | 24.64 | 148.4 | [v-1575](https://edgefirst.studio/public/validation/v-1575/details?mode=charts) |
+| Nano | Qualcomm Dragonwing IQ-9075 — Hexagon NPU (INT8) | ONNX · QNN EP | Logical | 49.21% | 28.07% | -1.17 | 16.83 | 209.2 | [v-156f](https://edgefirst.studio/public/validation/v-156f/details?mode=charts) |
+| Nano | Qualcomm Dragonwing IQ-9075 — Hexagon NPU (INT16) | ONNX · QNN EP | Smart | 51.34% | 29.18% | -0.06 | 31.32 | 131.8 | [v-1578](https://edgefirst.studio/public/validation/v-1578/details?mode=charts) |
+| Small | ONNX FP32 (AWS Graviton · 4-core) | ONNX | — | 60.07% | 34.82% | -0.01 | 1151.13 | 3.5 | [v-e5e](https://edgefirst.studio/public/validation/v-e5e/details?mode=charts) · 2026 Q2 |
+| Small | ONNX FP32 (AWS Graviton4 · 48-core) | ONNX | — | 60.07% | 34.82% | -0.01 | 311.98 | 42.7 | [v-e92](https://edgefirst.studio/public/validation/v-e92/details?mode=charts) · 2026 Q2 |
+| Small | ONNX FP32 (AWS Graviton4 · 8-core) | ONNX | — | 60.07% | 34.82% | -0.01 | 930.42 | 8.6 | [v-e8c](https://edgefirst.studio/public/validation/v-e8c/details?mode=charts) · 2026 Q2 |
+| Small | ONNX FP32 (Intel Core i9-13900F · 32-core) | ONNX | — | 60.03% | 34.83% | +0.00 | 159.33 | 19.8 | [v-a50](https://edgefirst.studio/public/validation/v-a50/details?mode=charts) · 2026 Q2 |
+| Small | ONNX FP32 (Intel Xeon Platinum 8488C · 24-core) | ONNX | — | 60.03% | 34.83% | +0.00 | 284.75 | 48.8 | [v-e94](https://edgefirst.studio/public/validation/v-e94/details?mode=charts) · 2026 Q2 |
+| Small | ONNX FP32 (Intel Xeon Platinum 8488C · 4-core) | ONNX | — | 60.03% | 34.83% | +0.00 | 638.96 | 12.8 | [v-e7e](https://edgefirst.studio/public/validation/v-e7e/details?mode=charts) · 2026 Q2 |
+| Small | ONNX FP32 (CUDA · A10G) | ONNX | — | 60.02% | 34.83% | ref | 35.30 | 137.3 | [v-e91](https://edgefirst.studio/public/validation/v-e91/details?mode=charts) · 2026 Q2 |
+| Small | ONNX FP32 (CUDA · RTX 4060) | ONNX | — | 60.02% | 34.83% | +0.00 | 27.76 | 151.7 | [v-a92](https://edgefirst.studio/public/validation/v-a92/details?mode=charts) · 2026 Q2 |
+| Small | ONNX FP16 (CUDA · RTX 4060) | ONNX | — | 60.03% | 34.83% | +0.00 | 22.06 | 222.5 | [v-aa7](https://edgefirst.studio/public/validation/v-aa7/details?mode=charts) · 2026 Q2 |
+| Small | ONNX FP32 (CUDA · RTX 3090) | ONNX | — | 60.32% | 35.03% | ref | 26.44 | 191.7 | [v-194e](https://edgefirst.studio/public/validation/v-194e/details?mode=charts) |
+| Small | ONNX FP16 (CUDA · RTX 3090) | ONNX | — | 60.33% | 35.06% | +0.03 | 21.28 | 263.6 | [v-19b3](https://edgefirst.studio/public/validation/v-19b3/details?mode=charts) |
+| Small | Apple M2 Max — CoreML Neural Engine (FP16) | ONNX | — | 59.62% | 34.65% | -0.18 | 11.48 | 261.8 | [v-9e9](https://edgefirst.studio/public/validation/v-9e9/details?mode=charts) · 2026 Q2 |
+| Small | Apple M2 Max — CoreML Metal GPU (FP16) | ONNX | — | 59.61% | 34.58% | -0.25 | 26.99 | 133.8 | [v-9e8](https://edgefirst.studio/public/validation/v-9e8/details?mode=charts) · 2026 Q2 |
+| Small | Apple M2 Max — CoreML CPU (FP16) | ONNX | — | 59.57% | 34.54% | -0.29 | 46.58 | 47.0 | [v-9e7](https://edgefirst.studio/public/validation/v-9e7/details?mode=charts) · 2026 Q2 |
+| Small | Apple iPhone 17 Pro — CoreML Neural Engine (FP16) | ONNX | — | 59.63% | 34.62% | -0.21 | 27.26 | 116.9 | [v-f6d](https://edgefirst.studio/public/validation/v-f6d/details?mode=charts) · 2026 Q2 |
+| Small | Apple iPhone 17 Pro — CoreML Metal GPU (FP16) | ONNX | — | 59.61% | 34.58% | -0.25 | 41.46 | 60.7 | [v-f6c](https://edgefirst.studio/public/validation/v-f6c/details?mode=charts) · 2026 Q2 |
+| Small | Apple iPhone 17 Pro — CoreML CPU (FP16) | ONNX | — | 59.56% | 34.55% | -0.28 | 82.71 | 27.2 | [v-f6b](https://edgefirst.studio/public/validation/v-f6b/details?mode=charts) · 2026 Q2 |
+| Small | Apple iPhone 15 Pro — CoreML Neural Engine (FP16) | ONNX | — | 59.63% | 34.63% | -0.20 | 37.03 | 102.0 | [v-f31](https://edgefirst.studio/public/validation/v-f31/details?mode=charts) · 2026 Q2 |
+| Small | Apple iPhone 15 Pro — CoreML Metal GPU (FP16) | ONNX | — | 59.62% | 34.59% | -0.24 | 83.77 | 26.0 | [v-f34](https://edgefirst.studio/public/validation/v-f34/details?mode=charts) · 2026 Q2 |
+| Small | Apple iPhone 15 Pro — CoreML CPU (FP16) | ONNX | — | 59.57% | 34.55% | -0.28 | 96.40 | 23.7 | [v-f02](https://edgefirst.studio/public/validation/v-f02/details?mode=charts) · 2026 Q2 |
+| Small | Samsung Galaxy S26 Ultra — Qualcomm Hexagon NPU (INT8) | ONNX · QNN EP | Smart | 58.59% | 33.71% | -1.12 | 30.71 | 117.2 | [v-10a0](https://edgefirst.studio/public/validation/v-10a0/details?mode=charts) · 2026 Q2 |
+| Small | Samsung Galaxy S26 Ultra — Qualcomm Hexagon NPU (INT8) | ONNX · QNN EP | Logical | 45.32% | 21.43% | -13.40 | 11.09 | 343.0 | [v-10a1](https://edgefirst.studio/public/validation/v-10a1/details?mode=charts) · 2026 Q2 |
+| Small | Samsung Galaxy S26 Ultra — Qualcomm Hexagon NPU (INT16) | ONNX · QNN EP | Smart | 59.48% | 34.40% | -0.43 | 40.59 | 93.6 | [v-109e](https://edgefirst.studio/public/validation/v-109e/details?mode=charts) · 2026 Q2 |
+| Small | Samsung Galaxy S26 Ultra — Qualcomm Hexagon NPU (INT16) | ONNX · QNN EP | Logical | 58.47% | 12.41% | -22.42 | 26.22 | 144.6 | [v-109f](https://edgefirst.studio/public/validation/v-109f/details?mode=charts) · 2026 Q2 |
+| Small | Samsung Galaxy S26 Ultra — CPU | LiteRT | Smart | 58.89% | 34.06% | -0.77 | 209.91 | 31.9 | [v-fe7](https://edgefirst.studio/public/validation/v-fe7/details?mode=charts) · 2026 Q2 |
+| Small | NXP i.MX 8M Plus + VeriSilicon NPU (Ezurio) — latency | LiteRT | Logical | 57.14% | 33.39% | -1.64 | 404.68 | 4.1 | [v-18e8](https://edgefirst.studio/public/validation/v-18e8/details?mode=charts) |
+| Small | NXP i.MX 8M Plus + VeriSilicon NPU (FRDM) | LiteRT | Combined | 45.88% | 27.30% | -7.73 | 168.05 | 6.3 | [v-18ce](https://edgefirst.studio/public/validation/v-18ce/details?mode=charts) |
+| Small | NXP i.MX 8M Plus + VeriSilicon NPU (FRDM) | LiteRT | Logical | 57.14% | 33.39% | -1.64 | 173.16 | 6.2 | [v-18e6](https://edgefirst.studio/public/validation/v-18e6/details?mode=charts) |
+| Small | NXP i.MX 8M Plus + VeriSilicon NPU (FRDM) | LiteRT | Smart | 59.36% | 34.34% | -0.69 | 193.70 | 6.5 | [v-18f7](https://edgefirst.studio/public/validation/v-18f7/details?mode=charts) |
+| Small | NXP i.MX 8M Plus + VeriSilicon NPU (Phytec) — latency | LiteRT | Logical | 57.05% | 33.37% | -1.66 | 178.05 | 6.4 | [v-18da](https://edgefirst.studio/public/validation/v-18da/details?mode=charts) |
+| Small | NXP i.MX 8M Plus + VeriSilicon NPU (Verdin) — latency | LiteRT | Logical | 57.14% | 33.39% | -1.64 | 178.25 | 6.1 | [v-18db](https://edgefirst.studio/public/validation/v-18db/details?mode=charts) |
+| Small | NXP i.MX 8M Plus + VeriSilicon NPU (Verdin) — throughput | LiteRT | Smart | 58.45% | 33.86% | -0.97 | 209.55 | 6.3 | [v-c92](https://edgefirst.studio/public/validation/v-c92/details?mode=charts) · 2026 Q2 |
+| Small | NXP i.MX 95 + eIQ Neutron NPU (Ezurio) — latency | LiteRT | Smart | 59.39% | 34.34% | -0.69 | 85.97 | 21.6 | [v-1678](https://edgefirst.studio/public/validation/v-1678/details?mode=charts) |
+| Small | NXP i.MX 95 + eIQ Neutron NPU (Ezurio) — throughput | LiteRT | Smart | 59.39% | 34.34% | -0.69 | 214.45 | 23.3 | [v-1905](https://edgefirst.studio/public/validation/v-1905/details?mode=charts) |
+| Small | NXP i.MX 95 + eIQ Neutron NPU (EVK) — latency | LiteRT | Smart | 59.39% | 34.35% | -0.68 | 85.80 | 21.7 | [v-1827](https://edgefirst.studio/public/validation/v-1827/details?mode=charts) |
+| Small | NXP i.MX 95 + eIQ Neutron NPU (EVK) — latency | LiteRT | Logical | 57.23% | 33.40% | -1.63 | 89.75 | 13.0 | [v-1820](https://edgefirst.studio/public/validation/v-1820/details?mode=charts) |
+| Small | NXP i.MX 95 + eIQ Neutron NPU (EVK) — latency | LiteRT | Combined | 46.04% | 27.46% | -7.57 | 90.92 | 12.4 | [v-1818](https://edgefirst.studio/public/validation/v-1818/details?mode=charts) |
+| Small | NXP i.MX 95 + eIQ Neutron NPU (EVK) — throughput | LiteRT | Smart | 59.39% | 34.35% | -0.68 | 212.58 | 23.6 | [v-182a](https://edgefirst.studio/public/validation/v-182a/details?mode=charts) |
+| Small | NXP i.MX 95 + eIQ Neutron NPU (EVK) — throughput | LiteRT | Logical | 57.23% | 33.40% | -1.63 | 194.35 | 22.9 | [v-1823](https://edgefirst.studio/public/validation/v-1823/details?mode=charts) |
+| Small | NXP i.MX 95 + eIQ Neutron NPU (EVK) — throughput | LiteRT | Combined | 46.04% | 27.46% | -7.57 | 178.41 | 23.9 | [v-181d](https://edgefirst.studio/public/validation/v-181d/details?mode=charts) |
+| Small | NXP i.MX 95 + eIQ Neutron NPU (Phytec) — throughput | LiteRT | Smart | 59.39% | 34.35% | -0.68 | 211.87 | 23.6 | [v-1780](https://edgefirst.studio/public/validation/v-1780/details?mode=charts) |
+| Small | NXP i.MX 95 + eIQ Neutron NPU (Verdin) — latency | LiteRT | Smart | 59.39% | 34.34% | -0.69 | 88.66 | 21.3 | [v-1679](https://edgefirst.studio/public/validation/v-1679/details?mode=charts) |
+| Small | NXP i.MX 95 + eIQ Neutron NPU (Verdin) — latency | LiteRT | Logical | 45.84% | 27.17% | -7.66 | 93.41 | 12.5 | [v-bdd](https://edgefirst.studio/public/validation/v-bdd/details?mode=charts) · 2026 Q2 |
+| Small | NXP i.MX 95 + eIQ Neutron NPU (Verdin) — throughput | LiteRT | Smart | 59.39% | 34.34% | -0.69 | 218.15 | 23.1 | [v-191e](https://edgefirst.studio/public/validation/v-191e/details?mode=charts) |
+| Small | NXP Ara240 (FRDM) — latency | Ara DVM | Smart | 58.29% | 33.76% | -1.27 | 42.22 | 50.4 | [v-1507](https://edgefirst.studio/public/validation/v-1507/details?mode=charts) |
+| Small | NXP Ara240 (FRDM) — throughput | Ara DVM | Smart | 58.29% | 33.76% | -1.27 | 50.14 | 63.5 | [v-18e2](https://edgefirst.studio/public/validation/v-18e2/details?mode=charts) |
+| Small | Raspberry Pi 5 + Hailo-8L NPU | Hailo HEF | — | 59.01% | 33.94% | -1.09 | 54.06 | 22.7 | [v-154d](https://edgefirst.studio/public/validation/v-154d/details?mode=charts) |
+| Small | NVIDIA Jetson Orin Nano (TensorRT FP16) | TensorRT | — | 60.32% | 35.03% | +0.00 | 68.77 | 82.2 | [v-1570](https://edgefirst.studio/public/validation/v-1570/details?mode=charts) |
+| Small | NVIDIA Jetson Orin Nano (TensorRT INT8) | TensorRT | — | 59.13% | 34.17% | -0.86 | 62.26 | 101.9 | [v-1577](https://edgefirst.studio/public/validation/v-1577/details?mode=charts) |
+| Small | Qualcomm Dragonwing IQ-9075 — Hexagon NPU (INT8) | ONNX · QNN EP | Smart | 59.71% | 34.74% | -0.29 | 22.55 | 178.4 | [v-15a0](https://edgefirst.studio/public/validation/v-15a0/details?mode=charts) |
+| Small | Qualcomm Dragonwing IQ-9075 — Hexagon NPU (INT8) | ONNX · QNN EP | Logical | 57.70% | 33.84% | -1.19 | 17.69 | 249.3 | [v-159b](https://edgefirst.studio/public/validation/v-159b/details?mode=charts) |
+| Small | Qualcomm Dragonwing IQ-9075 — Hexagon NPU (INT16) | ONNX · QNN EP | Smart | 60.19% | 35.00% | -0.03 | 32.82 | 148.0 | [v-15a2](https://edgefirst.studio/public/validation/v-15a2/details?mode=charts) |
+| Medium | ONNX FP32 (AWS Graviton · 4-core) | ONNX | — | 65.07% | 38.13% | -0.01 | 2829.07 | 1.4 | [v-e5f](https://edgefirst.studio/public/validation/v-e5f/details?mode=charts) · 2026 Q2 |
+| Medium | ONNX FP32 (AWS Graviton4 · 48-core) | ONNX | — | 65.07% | 38.13% | -0.01 | 737.76 | 17.6 | [v-e9e](https://edgefirst.studio/public/validation/v-e9e/details?mode=charts) · 2026 Q2 |
+| Medium | ONNX FP32 (AWS Graviton4 · 8-core) | ONNX | — | 65.07% | 38.13% | -0.01 | 2253.46 | 3.6 | [v-e76](https://edgefirst.studio/public/validation/v-e76/details?mode=charts) · 2026 Q2 |
+| Medium | ONNX FP32 (Intel Core i9-13900F · 32-core) | ONNX | — | 65.08% | 38.14% | +0.00 | 340.38 | 8.9 | [v-a57](https://edgefirst.studio/public/validation/v-a57/details?mode=charts) · 2026 Q2 |
+| Medium | ONNX FP32 (Intel Xeon Platinum 8488C · 24-core) | ONNX | — | 65.08% | 38.14% | +0.00 | 618.47 | 21.2 | [v-e99](https://edgefirst.studio/public/validation/v-e99/details?mode=charts) · 2026 Q2 |
+| Medium | ONNX FP32 (Intel Xeon Platinum 8488C · 4-core) | ONNX | — | 65.08% | 38.14% | +0.00 | 1443.22 | 5.6 | [v-e81](https://edgefirst.studio/public/validation/v-e81/details?mode=charts) · 2026 Q2 |
+| Medium | ONNX FP32 (CUDA · A10G) | ONNX | — | 65.08% | 38.14% | ref | 44.40 | 112.7 | [v-e8f](https://edgefirst.studio/public/validation/v-e8f/details?mode=charts) · 2026 Q2 |
+| Medium | ONNX FP32 (CUDA · RTX 4060) | ONNX | — | 65.08% | 38.15% | +0.01 | 62.02 | 62.1 | [v-a99](https://edgefirst.studio/public/validation/v-a99/details?mode=charts) · 2026 Q2 |
+| Medium | ONNX FP16 (CUDA · RTX 4060) | ONNX | — | 65.06% | 38.15% | +0.01 | 37.71 | 118.2 | [v-aae](https://edgefirst.studio/public/validation/v-aae/details?mode=charts) · 2026 Q2 |
+| Medium | ONNX FP32 (CUDA · RTX 3090) | ONNX | — | 65.52% | 38.62% | ref | 34.66 | 143.5 | [v-1951](https://edgefirst.studio/public/validation/v-1951/details?mode=charts) |
+| Medium | ONNX FP16 (CUDA · RTX 3090) | ONNX | — | 65.53% | 38.62% | +0.00 | 25.60 | 209.0 | [v-19b5](https://edgefirst.studio/public/validation/v-19b5/details?mode=charts) |
+| Medium | Apple M2 Max — CoreML Neural Engine (FP16) | ONNX | — | 63.08% | 37.10% | -1.04 | 21.17 | 111.3 | [v-72e](https://edgefirst.studio/public/validation/v-72e/details?mode=charts) · 2026 Q2 |
+| Medium | Apple M2 Max — CoreML Metal GPU (FP16) | ONNX | — | 64.09% | 37.54% | -0.60 | 55.85 | 58.0 | [v-9df](https://edgefirst.studio/public/validation/v-9df/details?mode=charts) · 2026 Q2 |
+| Medium | Apple M2 Max — CoreML CPU (FP16) | ONNX | — | 64.07% | 37.51% | -0.63 | 83.78 | 25.1 | [v-9de](https://edgefirst.studio/public/validation/v-9de/details?mode=charts) · 2026 Q2 |
+| Medium | Apple iPhone 17 Pro — CoreML Neural Engine (FP16) | ONNX | — | 64.05% | 37.62% | -0.52 | 32.78 | 73.9 | [v-f74](https://edgefirst.studio/public/validation/v-f74/details?mode=charts) · 2026 Q2 |
+| Medium | Apple iPhone 17 Pro — CoreML Metal GPU (FP16) | ONNX | — | 64.09% | 37.55% | -0.59 | 74.28 | 29.0 | [v-f76](https://edgefirst.studio/public/validation/v-f76/details?mode=charts) · 2026 Q2 |
+| Medium | Apple iPhone 17 Pro — CoreML CPU (FP16) | ONNX | — | 64.06% | 37.51% | -0.63 | 174.04 | 12.1 | [v-f75](https://edgefirst.studio/public/validation/v-f75/details?mode=charts) · 2026 Q2 |
+| Medium | Apple iPhone 15 Pro — CoreML Neural Engine (FP16) | ONNX | — | 64.05% | 37.62% | -0.52 | 42.71 | 57.9 | [v-f43](https://edgefirst.studio/public/validation/v-f43/details?mode=charts) · 2026 Q2 |
+| Medium | Apple iPhone 15 Pro — CoreML Metal GPU (FP16) | ONNX | — | 64.10% | 37.53% | -0.61 | 193.21 | 10.8 | [v-f4e](https://edgefirst.studio/public/validation/v-f4e/details?mode=charts) · 2026 Q2 |
+| Medium | Apple iPhone 15 Pro — CoreML CPU (FP16) | ONNX | — | 64.06% | 37.53% | -0.61 | 195.47 | 10.9 | [v-f05](https://edgefirst.studio/public/validation/v-f05/details?mode=charts) · 2026 Q2 |
+| Medium | Samsung Galaxy S26 Ultra — Qualcomm Hexagon NPU (INT8) | ONNX · QNN EP | Smart | 63.95% | 37.29% | -0.85 | 35.70 | 109.8 | [v-10a4](https://edgefirst.studio/public/validation/v-10a4/details?mode=charts) · 2026 Q2 |
+| Medium | Samsung Galaxy S26 Ultra — Qualcomm Hexagon NPU (INT8) | ONNX · QNN EP | Logical | 47.02% | 24.48% | -13.66 | 14.59 | 258.3 | [v-10a5](https://edgefirst.studio/public/validation/v-10a5/details?mode=charts) · 2026 Q2 |
+| Medium | Samsung Galaxy S26 Ultra — Qualcomm Hexagon NPU (INT16) | ONNX · QNN EP | Smart | 64.77% | 37.82% | -0.32 | 46.64 | 80.8 | [v-10a2](https://edgefirst.studio/public/validation/v-10a2/details?mode=charts) · 2026 Q2 |
+| Medium | Samsung Galaxy S26 Ultra — Qualcomm Hexagon NPU (INT16) | ONNX · QNN EP | Logical | 63.71% | 13.22% | -24.92 | 33.54 | 108.8 | [v-10a3](https://edgefirst.studio/public/validation/v-10a3/details?mode=charts) · 2026 Q2 |
+| Medium | Samsung Galaxy S26 Ultra — CPU | LiteRT | Smart | 63.62% | 37.20% | -0.94 | 457.07 | 13.7 | [v-101f](https://edgefirst.studio/public/validation/v-101f/details?mode=charts) · 2026 Q2 |
+| Medium | NXP i.MX 8M Plus + VeriSilicon NPU (Ezurio) — latency | LiteRT | Logical | 61.92% | 36.78% | -1.84 | 563.39 | 3.6 | [v-194b](https://edgefirst.studio/public/validation/v-194b/details?mode=charts) |
+| Medium | NXP i.MX 8M Plus + VeriSilicon NPU (FRDM) | LiteRT | Combined | 48.09% | 29.17% | -9.45 | 290.98 | 3.5 | [v-1906](https://edgefirst.studio/public/validation/v-1906/details?mode=charts) |
+| Medium | NXP i.MX 8M Plus + VeriSilicon NPU (FRDM) | LiteRT | Logical | 61.92% | 36.78% | -1.84 | 295.83 | 3.5 | [v-1963](https://edgefirst.studio/public/validation/v-1963/details?mode=charts) |
+| Medium | NXP i.MX 8M Plus + VeriSilicon NPU (FRDM) | LiteRT | Smart | 64.63% | 38.04% | -0.58 | 316.34 | 3.6 | [v-19c6](https://edgefirst.studio/public/validation/v-19c6/details?mode=charts) |
+| Medium | NXP i.MX 8M Plus + VeriSilicon NPU (Phytec) — latency | LiteRT | Logical | 61.75% | 36.67% | -1.95 | 303.00 | 3.6 | [v-1911](https://edgefirst.studio/public/validation/v-1911/details?mode=charts) |
+| Medium | NXP i.MX 8M Plus + VeriSilicon NPU (Verdin) — latency | LiteRT | Logical | 61.86% | 36.78% | -1.84 | 304.38 | 3.5 | [v-191c](https://edgefirst.studio/public/validation/v-191c/details?mode=charts) |
+| Medium | NXP i.MX 8M Plus + VeriSilicon NPU (Verdin) — throughput | LiteRT | Smart | 62.77% | 36.63% | -1.51 | 331.78 | 3.6 | [v-ca6](https://edgefirst.studio/public/validation/v-ca6/details?mode=charts) · 2026 Q2 |
+| Medium | NXP i.MX 95 + eIQ Neutron NPU (Ezurio) — latency | LiteRT | Smart | 64.61% | 37.98% | -0.64 | 155.70 | 9.2 | [v-1687](https://edgefirst.studio/public/validation/v-1687/details?mode=charts) |
+| Medium | NXP i.MX 95 + eIQ Neutron NPU (Ezurio) — throughput | LiteRT | Smart | 64.61% | 37.98% | -0.64 | 463.68 | 9.6 | [v-191b](https://edgefirst.studio/public/validation/v-191b/details?mode=charts) |
+| Medium | NXP i.MX 95 + eIQ Neutron NPU (EVK) — latency | LiteRT | Smart | 64.55% | 37.94% | -0.68 | 143.45 | 9.2 | [v-1842](https://edgefirst.studio/public/validation/v-1842/details?mode=charts) |
+| Medium | NXP i.MX 95 + eIQ Neutron NPU (EVK) — latency | LiteRT | Logical | 61.85% | 36.78% | -1.84 | 153.26 | 7.1 | [v-183b](https://edgefirst.studio/public/validation/v-183b/details?mode=charts) |
+| Medium | NXP i.MX 95 + eIQ Neutron NPU (EVK) — latency | LiteRT | Combined | 48.26% | 29.31% | -9.31 | 152.89 | 7.0 | [v-1830](https://edgefirst.studio/public/validation/v-1830/details?mode=charts) |
+| Medium | NXP i.MX 95 + eIQ Neutron NPU (EVK) — throughput | LiteRT | Smart | 64.55% | 37.94% | -0.68 | 451.71 | 9.6 | [v-1844](https://edgefirst.studio/public/validation/v-1844/details?mode=charts) |
+| Medium | NXP i.MX 95 + eIQ Neutron NPU (EVK) — throughput | LiteRT | Logical | 61.85% | 36.78% | -1.84 | 444.79 | 9.4 | [v-183e](https://edgefirst.studio/public/validation/v-183e/details?mode=charts) |
+| Medium | NXP i.MX 95 + eIQ Neutron NPU (EVK) — throughput | LiteRT | Combined | 48.26% | 29.31% | -9.31 | 428.11 | 9.6 | [v-1839](https://edgefirst.studio/public/validation/v-1839/details?mode=charts) |
+| Medium | NXP i.MX 95 + eIQ Neutron NPU (Phytec) — throughput | LiteRT | Smart | 64.55% | 37.94% | -0.68 | 451.25 | 9.6 | [v-1789](https://edgefirst.studio/public/validation/v-1789/details?mode=charts) |
+| Medium | NXP i.MX 95 + eIQ Neutron NPU (Verdin) — latency | LiteRT | Smart | 64.61% | 37.98% | -0.64 | 156.06 | 9.2 | [v-1689](https://edgefirst.studio/public/validation/v-1689/details?mode=charts) |
+| Medium | NXP i.MX 95 + eIQ Neutron NPU (Verdin) — latency | LiteRT | Logical | 47.18% | 28.34% | -9.80 | 167.73 | 6.5 | [v-beb](https://edgefirst.studio/public/validation/v-beb/details?mode=charts) · 2026 Q2 |
+| Medium | NXP i.MX 95 + eIQ Neutron NPU (Verdin) — throughput | LiteRT | Smart | 64.61% | 37.98% | -0.64 | 465.43 | 9.5 | [v-192c](https://edgefirst.studio/public/validation/v-192c/details?mode=charts) |
+| Medium | NXP Ara240 (FRDM) — latency | Ara DVM | Smart | 61.63% | 36.02% | -2.60 | 54.85 | 29.6 | [v-1516](https://edgefirst.studio/public/validation/v-1516/details?mode=charts) |
+| Medium | NXP Ara240 (FRDM) — throughput | Ara DVM | Smart | 61.61% | 36.00% | -2.62 | 48.03 | 37.9 | [v-18e5](https://edgefirst.studio/public/validation/v-18e5/details?mode=charts) |
+| Medium | Raspberry Pi 5 + Hailo-8L NPU | Hailo HEF | — | 64.79% | 38.25% | -0.37 | 84.72 | 13.3 | [v-1564](https://edgefirst.studio/public/validation/v-1564/details?mode=charts) |
+| Medium | NVIDIA Jetson Orin Nano (TensorRT FP16) | TensorRT | — | 65.51% | 38.62% | +0.00 | 84.58 | 58.2 | [v-1579](https://edgefirst.studio/public/validation/v-1579/details?mode=charts) |
+| Medium | NVIDIA Jetson Orin Nano (TensorRT INT8) | TensorRT | — | 64.35% | 37.80% | -0.82 | 64.09 | 93.7 | [v-157e](https://edgefirst.studio/public/validation/v-157e/details?mode=charts) |
+| Medium | Qualcomm Dragonwing IQ-9075 — Hexagon NPU (INT8) | ONNX · QNN EP | Smart | 64.93% | 38.24% | -0.38 | 27.50 | 175.9 | [v-15dd](https://edgefirst.studio/public/validation/v-15dd/details?mode=charts) |
+| Medium | Qualcomm Dragonwing IQ-9075 — Hexagon NPU (INT8) | ONNX · QNN EP | Logical | 62.44% | 37.35% | -1.27 | 22.92 | 201.0 | [v-15d3](https://edgefirst.studio/public/validation/v-15d3/details?mode=charts) |
+| Medium | Qualcomm Dragonwing IQ-9075 — Hexagon NPU (INT16) | ONNX · QNN EP | Smart | 65.44% | 38.47% | -0.15 | 37.91 | 87.7 | [v-15e8](https://edgefirst.studio/public/validation/v-15e8/details?mode=charts) |
 
 
 
@@ -251,9 +278,9 @@ These results are produced by the EdgeFirst on-target validation pipeline:
 2. **EdgeFirst Studio** receives the metrics, predictions, charts and trace, and publishes them as the `v-XXXX` validation session every row above cites. Studio is where results are published, compared and browsed — it is not where they are computed.
 3. **EdgeFirst HAL** ([open source](https://github.com/EdgeFirstAI/hal)) provides the hardware-accelerated preprocessing and post-decoding primitives used at both validation and deployment time, so the timings measured here reflect the same accelerated paths a production runtime would take.
 
-Inference latency is reported as the on-accelerator inference time. End-to-end latency is the sequential per-image latency across the compute pipeline — preprocessing, inference, and postprocessing; image acquisition (file or camera load and JPEG decode) overlaps these stages and is excluded from this figure.
+End-to-end latency runs from the moment a frame is available to the moment its results are ready — preprocessing, inference, and postprocessing. Image acquisition (camera capture, or file load and JPEG decode), mask encoding and writing results to disk are excluded. The per-stage breakdown of every row is on its linked Studio session.
 
-Two throughput figures are reported. **Realized FPS** is the *measured* steady-state rate at which final results are emitted, measured directly from the profiler's per-frame result-emission timestamps over the steady-state stream — trace-independent; the Perfetto trace's own FPS is used only as a fallback on sessions where that scalar isn't available. It is the true, priority number and generally exceeds `1000 / end-to-end` because the runtime overlaps stages across frames. **Core-throughput ceiling (FPS)** is the accelerator's **core** ceiling — `1000 / device-compute-time`, i.e. the throughput if the accelerator were the *only* bottleneck. It is taken from the isolated device-compute stage (on transfer-split runtimes the trace separates host↔device transfers from device compute), which is load-independent — unlike the host capture/preprocess service times, whose measured cost inflates under pipeline backpressure (the same 5000 JPEGs cost ~7.8 ms/frame serialized but far more under throughput backpressure), so the slowest-stage figure would *understate* a fast accelerator. It is a possibly-achievable ceiling, not a measured result: reaching it depends on the deployment pipeline. A validation run decodes a JPEG per image and evaluates at a `0.001` confidence threshold (to capture every detection for mAP), both of which load the host and postprocess stages; a production camera pipeline (no JPEG decode) at a deployment threshold of `0.25–0.75` (far fewer candidate boxes through NMS) moves realized throughput toward the core-throughput ceiling.
+**Realized FPS** is the *measured* steady-state rate at which final results are emitted, measured directly from the profiler's per-frame result-emission timestamps over the steady-state stream — trace-independent; the Perfetto trace's own FPS is used only as a fallback on sessions where that scalar isn't available. It generally exceeds `1000 / end-to-end` because the runtime overlaps stages across frames.
 
 See [EdgeFirst Studio](https://edgefirst.studio) for the full validation pipeline.
 
@@ -263,19 +290,137 @@ See [EdgeFirst Studio](https://edgefirst.studio) for the full validation pipelin
 
 Artifacts are organized by deployment target. Each model file embeds the EdgeFirst `edgefirst.json` metadata (training session, dataset version, calibration artifact, converter chain) so a single file is sufficient for deployment — no sidecar configuration required.
 
-Browse and download every artifact from the **[repository file tree](https://huggingface.co/EdgeFirst/yolov8-seg/tree/main)**. Files are organized into per-target folders and follow the naming convention `yolov8{size}-seg-{precision}[-smart]{extension}`:
+Every file below is the artifact a validation session in the tables above measured. Files are organized into per-target folders and follow the naming convention `yolov8{size}-seg-{precision}[-smart|-logical][-end2end]{extension}`.
+
+Entries marked **2026 Q2** belong to rows carried from the 2026 Q2 edition: they download from the repository's `2026Q2` tag, the weights those rows were measured on. The same filename on `main` holds the current edition's artifact.
+
+<details open>
+<summary><strong>ONNX FP32</strong> — Any platform with ONNX Runtime.</summary>
+
+| Size | File | Variant | Download |
+|------|------|---------|----------|
+| Nano | `onnx/yolov8n-seg-fp16.onnx` | FP16 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/onnx/yolov8n-seg-fp16.onnx) |
+| Nano | `onnx/yolov8n-seg-fp16.onnx` | FP16 · 2026 Q2 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/2026Q2/onnx/yolov8n-seg-fp16.onnx) |
+| Nano | `onnx/yolov8n-seg-fp32.onnx` | FP32 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/onnx/yolov8n-seg-fp32.onnx) |
+| Nano | `onnx/yolov8n-seg-fp32.onnx` | FP32 · 2026 Q2 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/2026Q2/onnx/yolov8n-seg-fp32.onnx) |
+| Small | `onnx/yolov8s-seg-fp16.onnx` | FP16 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/onnx/yolov8s-seg-fp16.onnx) |
+| Small | `onnx/yolov8s-seg-fp16.onnx` | FP16 · 2026 Q2 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/2026Q2/onnx/yolov8s-seg-fp16.onnx) |
+| Small | `onnx/yolov8s-seg-fp32.onnx` | FP32 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/onnx/yolov8s-seg-fp32.onnx) |
+| Small | `onnx/yolov8s-seg-fp32.onnx` | FP32 · 2026 Q2 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/2026Q2/onnx/yolov8s-seg-fp32.onnx) |
+| Medium | `onnx/yolov8m-seg-fp16.onnx` | FP16 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/onnx/yolov8m-seg-fp16.onnx) |
+| Medium | `onnx/yolov8m-seg-fp16.onnx` | FP16 · 2026 Q2 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/2026Q2/onnx/yolov8m-seg-fp16.onnx) |
+| Medium | `onnx/yolov8m-seg-fp32.onnx` | FP32 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/onnx/yolov8m-seg-fp32.onnx) |
+| Medium | `onnx/yolov8m-seg-fp32.onnx` | FP32 · 2026 Q2 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/2026Q2/onnx/yolov8m-seg-fp32.onnx) |
+
+</details>
+
+<details>
+<summary><strong>TFLite INT8</strong> — CPU or NPU via runtime delegate (NXP i.MX 8M Plus VX Delegate).</summary>
+
+| Size | File | Variant | Download |
+|------|------|---------|----------|
+| Nano | `tflite/yolov8n-seg-int8-smart.tflite` | Smart · INT8 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/tflite/yolov8n-seg-int8-smart.tflite) |
+| Nano | `tflite/yolov8n-seg-int8-smart.tflite` | Smart · INT8 · 2026 Q2 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/2026Q2/tflite/yolov8n-seg-int8-smart.tflite) |
+| Nano | `tflite/yolov8n-seg-int8.tflite` | Logical · INT8 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/tflite/yolov8n-seg-int8.tflite) |
+| Small | `tflite/yolov8s-seg-int8-smart.tflite` | Smart · INT8 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/tflite/yolov8s-seg-int8-smart.tflite) |
+| Small | `tflite/yolov8s-seg-int8-smart.tflite` | Smart · INT8 · 2026 Q2 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/2026Q2/tflite/yolov8s-seg-int8-smart.tflite) |
+| Small | `tflite/yolov8s-seg-int8.tflite` | Logical · INT8 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/tflite/yolov8s-seg-int8.tflite) |
+| Medium | `tflite/yolov8m-seg-int8-smart.tflite` | Smart · INT8 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/tflite/yolov8m-seg-int8-smart.tflite) |
+| Medium | `tflite/yolov8m-seg-int8-smart.tflite` | Smart · INT8 · 2026 Q2 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/2026Q2/tflite/yolov8m-seg-int8-smart.tflite) |
+| Medium | `tflite/yolov8m-seg-int8.tflite` | Logical · INT8 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/tflite/yolov8m-seg-int8.tflite) |
+
+</details>
+
+<details>
+<summary><strong>NXP i.MX 95 (eIQ Neutron)</strong> — eIQ Neutron NPU optimized.</summary>
+
+| Size | File | Variant | Download |
+|------|------|---------|----------|
+| Nano | `imx95/yolov8n-seg-int8-smart.imx95.tflite` | Smart · INT8 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/imx95/yolov8n-seg-int8-smart.imx95.tflite) |
+| Nano | `imx95/yolov8n-seg-int8.imx95.tflite` | Logical · INT8 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/imx95/yolov8n-seg-int8.imx95.tflite) |
+| Nano | `imx95/yolov8n-seg-int8.imx95.tflite` | Logical · INT8 · 2026 Q2 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/2026Q2/imx95/yolov8n-seg-int8.imx95.tflite) |
+| Small | `imx95/yolov8s-seg-int8-smart.imx95.tflite` | Smart · INT8 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/imx95/yolov8s-seg-int8-smart.imx95.tflite) |
+| Small | `imx95/yolov8s-seg-int8.imx95.tflite` | Logical · INT8 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/imx95/yolov8s-seg-int8.imx95.tflite) |
+| Small | `imx95/yolov8s-seg-int8.imx95.tflite` | Logical · INT8 · 2026 Q2 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/2026Q2/imx95/yolov8s-seg-int8.imx95.tflite) |
+| Medium | `imx95/yolov8m-seg-int8-smart.imx95.tflite` | Smart · INT8 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/imx95/yolov8m-seg-int8-smart.imx95.tflite) |
+| Medium | `imx95/yolov8m-seg-int8.imx95.tflite` | Logical · INT8 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/imx95/yolov8m-seg-int8.imx95.tflite) |
+| Medium | `imx95/yolov8m-seg-int8.imx95.tflite` | Logical · INT8 · 2026 Q2 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/2026Q2/imx95/yolov8m-seg-int8.imx95.tflite) |
+
+</details>
+
+<details>
+<summary><strong>NXP Ara240</strong> — Ara240 DNPU compiled model.</summary>
+
+| Size | File | Variant | Download |
+|------|------|---------|----------|
+| Nano | `ara240/yolov8n-seg-int16.dvm` | Smart · INT16 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/ara240/yolov8n-seg-int16.dvm) |
+| Small | `ara240/yolov8s-seg-int16.dvm` | Smart · INT16 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/ara240/yolov8s-seg-int16.dvm) |
+| Medium | `ara240/yolov8m-seg-int16.dvm` | Smart · INT16 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/ara240/yolov8m-seg-int16.dvm) |
+
+</details>
+
+<details>
+<summary><strong>RPi5 + Hailo-8L (13 TOPS)</strong> — Hailo-8L 13 TOPS accelerator.</summary>
+
+| Size | File | Variant | Download |
+|------|------|---------|----------|
+| Nano | `hailo/yolov8n-seg-int8.hailo8l.hef` | INT8 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/hailo/yolov8n-seg-int8.hailo8l.hef) |
+| Small | `hailo/yolov8s-seg-int8.hailo8l.hef` | INT8 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/hailo/yolov8s-seg-int8.hailo8l.hef) |
+| Medium | `hailo/yolov8m-seg-int8.hailo8l.hef` | INT8 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/hailo/yolov8m-seg-int8.hailo8l.hef) |
+
+</details>
+
+<details>
+<summary><strong>NVIDIA Jetson (TensorRT)</strong> — Jetson FP16 and INT8 engines.</summary>
+
+| Size | File | Variant | Download |
+|------|------|---------|----------|
+| Nano | `jetson/yolov8n-seg-fp16.engine` | FP16 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/jetson/yolov8n-seg-fp16.engine) |
+| Nano | `jetson/yolov8n-seg-int8.engine` | INT8 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/jetson/yolov8n-seg-int8.engine) |
+| Small | `jetson/yolov8s-seg-fp16.engine` | FP16 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/jetson/yolov8s-seg-fp16.engine) |
+| Small | `jetson/yolov8s-seg-int8.engine` | INT8 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/jetson/yolov8s-seg-int8.engine) |
+| Medium | `jetson/yolov8m-seg-fp16.engine` | FP16 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/jetson/yolov8m-seg-fp16.engine) |
+| Medium | `jetson/yolov8m-seg-int8.engine` | INT8 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/jetson/yolov8m-seg-int8.engine) |
+
+</details>
+
+<details>
+<summary><strong>Qualcomm Snapdragon 8 Elite Gen 5 (Hexagon v81)</strong> — Offline-compiled Hexagon context binary, run through the ONNX Runtime QNN execution provider. Loads only on the Snapdragon 8 Elite Gen 5 (Hexagon v81).</summary>
+
+| Size | File | Variant | Download |
+|------|------|---------|----------|
+| Nano | `qnn/yolov8n-seg-int16-logical.htp-v81.qnn.onnx` | Logical · INT16 · 2026 Q2 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/2026Q2/qnn/yolov8n-seg-int16-logical.htp-v81.qnn.onnx) |
+| Nano | `qnn/yolov8n-seg-int16-smart.htp-v81.qnn.onnx` | Smart · INT16 · 2026 Q2 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/2026Q2/qnn/yolov8n-seg-int16-smart.htp-v81.qnn.onnx) |
+| Nano | `qnn/yolov8n-seg-int8-logical.htp-v81.qnn.onnx` | Logical · INT8 · 2026 Q2 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/2026Q2/qnn/yolov8n-seg-int8-logical.htp-v81.qnn.onnx) |
+| Nano | `qnn/yolov8n-seg-int8-smart.htp-v81.qnn.onnx` | Smart · INT8 · 2026 Q2 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/2026Q2/qnn/yolov8n-seg-int8-smart.htp-v81.qnn.onnx) |
+| Small | `qnn/yolov8s-seg-int16-logical.htp-v81.qnn.onnx` | Logical · INT16 · 2026 Q2 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/2026Q2/qnn/yolov8s-seg-int16-logical.htp-v81.qnn.onnx) |
+| Small | `qnn/yolov8s-seg-int16-smart.htp-v81.qnn.onnx` | Smart · INT16 · 2026 Q2 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/2026Q2/qnn/yolov8s-seg-int16-smart.htp-v81.qnn.onnx) |
+| Small | `qnn/yolov8s-seg-int8-logical.htp-v81.qnn.onnx` | Logical · INT8 · 2026 Q2 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/2026Q2/qnn/yolov8s-seg-int8-logical.htp-v81.qnn.onnx) |
+| Small | `qnn/yolov8s-seg-int8-smart.htp-v81.qnn.onnx` | Smart · INT8 · 2026 Q2 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/2026Q2/qnn/yolov8s-seg-int8-smart.htp-v81.qnn.onnx) |
+| Medium | `qnn/yolov8m-seg-int16-logical.htp-v81.qnn.onnx` | Logical · INT16 · 2026 Q2 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/2026Q2/qnn/yolov8m-seg-int16-logical.htp-v81.qnn.onnx) |
+| Medium | `qnn/yolov8m-seg-int16-smart.htp-v81.qnn.onnx` | Smart · INT16 · 2026 Q2 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/2026Q2/qnn/yolov8m-seg-int16-smart.htp-v81.qnn.onnx) |
+| Medium | `qnn/yolov8m-seg-int8-logical.htp-v81.qnn.onnx` | Logical · INT8 · 2026 Q2 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/2026Q2/qnn/yolov8m-seg-int8-logical.htp-v81.qnn.onnx) |
+| Medium | `qnn/yolov8m-seg-int8-smart.htp-v81.qnn.onnx` | Smart · INT8 · 2026 Q2 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/2026Q2/qnn/yolov8m-seg-int8-smart.htp-v81.qnn.onnx) |
+
+</details>
+
+<details>
+<summary><strong>Qualcomm Dragonwing IQ-9075 (Hexagon v73)</strong> — Offline-compiled Hexagon context binary, run through the ONNX Runtime QNN execution provider. Loads only on the Dragonwing IQ-9075 (Hexagon v73).</summary>
+
+| Size | File | Variant | Download |
+|------|------|---------|----------|
+| Nano | `qnn/yolov8n-seg-int16-smart.iq9075.qnn.onnx` | Smart · INT16 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/qnn/yolov8n-seg-int16-smart.iq9075.qnn.onnx) |
+| Nano | `qnn/yolov8n-seg-int8-logical.iq9075.qnn.onnx` | Logical · INT8 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/qnn/yolov8n-seg-int8-logical.iq9075.qnn.onnx) |
+| Nano | `qnn/yolov8n-seg-int8-smart.iq9075.qnn.onnx` | Smart · INT8 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/qnn/yolov8n-seg-int8-smart.iq9075.qnn.onnx) |
+| Small | `qnn/yolov8s-seg-int16-smart.iq9075.qnn.onnx` | Smart · INT16 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/qnn/yolov8s-seg-int16-smart.iq9075.qnn.onnx) |
+| Small | `qnn/yolov8s-seg-int8-logical.iq9075.qnn.onnx` | Logical · INT8 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/qnn/yolov8s-seg-int8-logical.iq9075.qnn.onnx) |
+| Small | `qnn/yolov8s-seg-int8-smart.iq9075.qnn.onnx` | Smart · INT8 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/qnn/yolov8s-seg-int8-smart.iq9075.qnn.onnx) |
+| Medium | `qnn/yolov8m-seg-int16-smart.iq9075.qnn.onnx` | Smart · INT16 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/qnn/yolov8m-seg-int16-smart.iq9075.qnn.onnx) |
+| Medium | `qnn/yolov8m-seg-int8-logical.iq9075.qnn.onnx` | Logical · INT8 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/qnn/yolov8m-seg-int8-logical.iq9075.qnn.onnx) |
+| Medium | `qnn/yolov8m-seg-int8-smart.iq9075.qnn.onnx` | Smart · INT8 | [Download](https://huggingface.co/EdgeFirst/yolov8-seg/resolve/main/qnn/yolov8m-seg-int8-smart.iq9075.qnn.onnx) |
+
+</details>
 
-| Target | Folder | Format |
-|--------|--------|--------|
-| ONNX FP32 | `onnx/` | `.onnx` |
-| TFLite INT8 | `tflite/` | `.tflite` |
-| NXP i.MX 95 (eIQ Neutron) | `imx95/` | `.imx95.tflite` |
-| NXP Ara240 | `ara240/` | `.dvm` |
-| RPi5 + Hailo-8L (13 TOPS) | `hailo/` | `.hailo8l.hef` |
-| NVIDIA Jetson (TensorRT) | `jetson/` | `.engine` |
-| Snapdragon (Qualcomm Hexagon HTP) | `qnn/` | `.htp-v81.qnn.onnx` |
-
-Each file embeds its `edgefirst.json` metadata (training session, dataset version, calibration artifact, converter chain), so a single download is sufficient for deployment — no sidecar configuration required.
 
 ---
 

ara240/yolov8m-seg-int16.dvm

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:edef956265449bee4f97d63518d991ca8dbd0bb5abdbb392fab7a55ad5b86893
-size 43027467
+oid sha256:daba458eada0d7c4dd9ef89d95f44763f2830890ce3b284b13bff3feba74887c
+size 43014101

ara240/yolov8n-seg-int16.dvm

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:74eeeae26bfe3cf65324dfebc473dc272674241340ace0878b95d79280657dc3
-size 6396850
+oid sha256:866e72fe9bea7a13d8c5cd4b01a7e4618e521dda588352ef44fe4c89f42677da
+size 6400390

ara240/yolov8s-seg-int16.dvm

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:28250c1a0a4bea27feb2d57bfb86e03a5ae948be12c25328fd1a1a320016698d
-size 18028583
+oid sha256:788a55b127fc3f6d14613c04b0956d8b4856185781bda272595369606e8cd038
+size 17997221

hailo/yolov8m-seg-int8.hailo8l.hef

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0313464192ce0c49d6a0f12e4323b8d43d6b8ea46fb0289bd77c99c39ac39048
-size 55127128
+oid sha256:99e3716129dc214f64272caa9ab70214b7460de7ef402148ce5aca5761fa5b3d
+size 55140130

hailo/yolov8n-seg-int8.hailo8l.hef

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fa535bb22496575e7653213913cf22a841f6d7c7f0d3fdd491a2e3e642f03883
-size 13548960
+oid sha256:2552ceb37f74cffcc1bec030f59e18ecf968575eae2f4969d3a083dd55b932d0
+size 13556133

hailo/yolov8s-seg-int8.hailo8l.hef

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b7c57aca531d9bbaa924c2f0bee342dfc2721ddb051a3a18c67f0cb923182af4
-size 30800574
+oid sha256:36098012c7e80379f1be1abe4506b6863270341044043032385526cee812920b
+size 30797570

imx95/yolov8m-seg-int8-smart.imx95.tflite

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c67bcd93a9ea118bdc2a09521e23b6df1232a0c9fc0d9928c51755ef74ecc72e
-size 26765088
+oid sha256:7296f04d1e3359a495f4783950f33214ef581aa68815d805d93254bccc376be6
+size 28543723

imx95/yolov8m-seg-int8.imx95.tflite

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b37ca0be47fa23211eddb114fb9a84387d903601f70cdd08a70664de0f1273ed
-size 26828275
+oid sha256:89bacc2ceaaacb18cedd44124941288fa8e9bcc6efe5a7bc4ca89aca76154859
+size 28614494

imx95/yolov8n-seg-int8-smart.imx95.tflite

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3f49578a60f58e97b826ecfbab86449aea57956673cb5ecd1ff89652917b3f0e
-size 3503301
+oid sha256:9d6b1fc18fc473765eb80d3e8a70001a628c1a5c02b1026adbe13d365927db43
+size 3526596

imx95/yolov8n-seg-int8.imx95.tflite

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e9a7c8cd59e6d774a5f6f6e416d1e17394d7e0d8237200f7e2b1675ba2b80c53
-size 3567010
+oid sha256:683d39b30357f9bc5b2f0ead13a09a24ff8f8c266079d1174696cb9a1d5ad5aa
+size 3597352

imx95/yolov8s-seg-int8-smart.imx95.tflite

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ecbc9e17bbd0a65e48bc1d1727df293f7b524579b2e5c131baace6c92117e90a
-size 11964739
+oid sha256:1751f75a18faebaff03c8dd88e81a00e58bec01d58c0f97df34b29b43931ef5d
+size 12345964

imx95/yolov8s-seg-int8.imx95.tflite

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1776871f7bcc265b4a70039c17af75ed7e4e5d1a8060795d7cb6c2a0de261d54
-size 12029154
+oid sha256:c7ab46551fe851f149f6b5c8ff3ff7566c4ba7aecbe3925596ae858b4e6420db
+size 12416715

jetson/yolov8m-seg-fp16.engine

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e6a100473fd30ca1366a80cfafbf3a036b9e9e580f57f17f652ddf7f3870dc8a
-size 58067391
+oid sha256:8b94c8596e02412a9541d526791bdd9b8bdf07ccfb92228e94e0cd48f584cd9f
+size 57955512

jetson/yolov8m-seg-int8.engine

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:13ce02639cc08df0dcb720bbbcd0f11529b021a9c6ac2f4957a90f1d92777f9d
+size 30814701

jetson/yolov8n-seg-fp16.engine

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ef17c08ded26169a39f811a8e5a9e7badc113bdc0140b0d85503dae3ba2f70ab
-size 9971604
+oid sha256:9fd794cbaead69f2c46756de7294a70aeec737b01cd73b73d2045732caccb21d
+size 9821141

jetson/yolov8n-seg-int8.engine

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d8d81e102f9177ecea9d3c55f65f4d1f8d31fb9d53917a834ef1f37edf30ca7
+size 5814882

jetson/yolov8s-seg-fp16.engine

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:20462f1f279292a6ae2e99bf7b1ac123dc9ae342131f39e1ee232c5cb83b6e4f
-size 26906446
+oid sha256:b9e6a9cd1a87a34053b21530837b25a269a33782984f571a91364873e65b14f2
+size 26858471

jetson/yolov8s-seg-int8.engine

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2535b3b806f73a067e2aca7b5406b427a573cd47f1d194887ff08830449b1c6b
+size 14260308

onnx/yolov8m-seg-fp16.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:179ca34f77d04e81ba9a2de683bc268f726f93e708bc56b97d0326557cd10b43
-size 54750813
+oid sha256:bfde118f6ae722e639b54b11415d91f80b6c78b243f5f2655d515e2d374390d2
+size 54751307

onnx/yolov8m-seg-fp32.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4b99bbf92d772dd6583b81351a36f0016629e2efadeddb125c28acadf49ce68b
-size 109368098
+oid sha256:2cb866cb1a515f9d7b981a408f1b8ea36a1adb36113dacb8e5765b3957ca270d
+size 109368472

onnx/yolov8n-seg-fp16.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fa6f8a1e814fbf429f52b1a66cfd42396e7e4f8ea50dadf0576c82462058f09d
-size 7000198
+oid sha256:1cf75ffa2b888820d3dafd99483afd7447d7391a6e82035a27431b83d1923dce
+size 7000692

onnx/yolov8n-seg-fp32.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1099cc6c6aea7fe8307ad4d3593b934a29523857e9aaa7e67f41cfd6934784cd
-size 13892965
+oid sha256:5348731f77e8d9352ce810d70a2eae60a6d5e63058e4662ec1d308cbcc2a2c25
+size 13893339

onnx/yolov8s-seg-fp16.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e01ebc31c152f7183531e9d3e623ba9b11130cdc6dca61f274ec8d9c508d47c1
-size 23812938
+oid sha256:256147340cade401a98a3ce2b21886a2d13875b5987e162ebb04dd55fa7d9c4b
+size 23813432

onnx/yolov8s-seg-fp32.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:992da5cb678d180265746f9aaa8b2b7d50e6e937ad0ae30bb3bea4bda88a871f
-size 47518105
+oid sha256:5aa59cbd61a41d60c2fed0d2de09a747e742570ad40079918d3fecf5bfceb87d
+size 47518479

qnn/yolov8m-seg-int16-smart.iq9075.qnn.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:edbde217011f1c8ca1f4ff438e974709c0de0d0f6a62655946e52425c190bcd0
+size 28981322

qnn/yolov8m-seg-int8-logical.iq9075.qnn.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c9698a1aea9cff3a3b982c0dc59f782a7f7b15e9105506b5b60c41b0e4b1e7b9
+size 28418200

qnn/yolov8m-seg-int8-smart.iq9075.qnn.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bdfcb63f3f6f04adb445617d76545f6d52df2259a41922e0f5f837c510e38388
+size 28362840

qnn/yolov8n-seg-int16-smart.iq9075.qnn.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:309cd1f1ae566b4090ce30ee0e2e325576a99471f0e718a01a4926ad5a2f8e6a
+size 4147922

qnn/yolov8n-seg-int8-logical.iq9075.qnn.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5eea0c05a9783ed517c804459a322b848e7026f844e55863fc53d734887c9d33
+size 4023072

qnn/yolov8n-seg-int8-smart.iq9075.qnn.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5181b09169272c365cadf0fbc110fda596b144dea51c5d34b12cbe59ad5c4d9a
+size 3955423

qnn/yolov8s-seg-int16-smart.iq9075.qnn.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:719cc0438dea2511f350b6461da4f82ca7cc40ad7e2b204668c264254cc73001
+size 12663505

qnn/yolov8s-seg-int8-logical.iq9075.qnn.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:562c53d48371971bd87503bde219261ca50a881c8104c388426ad9b391c1d027
+size 12432162

qnn/yolov8s-seg-int8-smart.iq9075.qnn.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:04f339b8041b8078c61d59e69c00e920014e5fb2c37abd23757d49baae052f31
+size 12368608

tflite/yolov8m-seg-int8-smart.tflite

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c1eafe79a6d44871645fc7884b2e965930e61d46794aaa13e53caf32036bc94d
-size 27863699
+oid sha256:4985957d47328f67dbbc7e9ed77200fcdc8847c656349fa599f6e9d7f0100426
+size 27864135

tflite/yolov8m-seg-int8.tflite

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:264dacbc848115119255dbab102680250251af0c8c02980c2291199204cc6a6c
-size 27913327
+oid sha256:b83e77b4cfdfbbf41d6a126f1504dbb44bca3df5653530299e163e89bc95a1d8
+size 27924987

tflite/yolov8n-seg-int8-smart.tflite

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4ac280c68bcd8fbc4019b713802a77fd81c6a0b2293f2007f6f64292c26db69c
-size 3660527
+oid sha256:988c115f75b453d2f00aa171c4542e7816b1ff0fe761f6ae112a033703d598c4
+size 3660951

tflite/yolov8n-seg-int8.tflite

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d436628e192c5d186ab46f4510417099ee61c3277cd7f4f17ff7dcabef02257a
-size 3712494
+oid sha256:0cbccb870cda3d89b3f600001809dfd64d4346cc76873dd9a87c7294f9b121e3
+size 3721788

tflite/yolov8s-seg-int8-smart.tflite

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a58359d9099abc74da7f409c9f2ba3df64afeefc6dac7dca5fb437812810ef33
-size 12199389
+oid sha256:ec43728035954441bba54cd2ff32ceca3ca9cd3819f2e122a686e14f06bba5d1
+size 12200005

tflite/yolov8s-seg-int8.tflite

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c7de64ffcd5aa9d93dcbb57b22d236551f96d7ee38b9e0ffb8b09b94f642fbfd
-size 12251358
+oid sha256:0461c8f5a3dd2fd1ed71584c454f68b0aeed5d402ddb9f979a00789b7774ebae
+size 12260832