Update model card for yolo11-det

README.md

@@ -28,29 +28,23 @@ model-index:
             value: 53.05
           - name: "mAP@0.5-0.95 (Nano ONNX FP32)"
             type: map
-            value: 37.82
+            value: 37.78
           - name: "mAP@0.5 (Small ONNX FP32)"
             type: map_50
-            value: 61.21
+            value: 61.19
           - name: "mAP@0.5-0.95 (Small ONNX FP32)"
             type: map
-            value: 44.91
-          - name: "mAP@0.5 (Medium ONNX FP32)"
-            type: map_50
-            value: 65.91
-          - name: "mAP@0.5-0.95 (Medium ONNX FP32)"
-            type: map
-            value: 49.59
+            value: 44.89
 ---
 
 # YOLO11 Detection — EdgeFirst Model Zoo
 
-YOLO11 Detection models trained on [COCO 2017](https://test.edgefirst.studio/public/projects/2839/home) (80 classes) and validated on real edge hardware through the EdgeFirst Profiler + Validator pipeline. Each row in the tables below cites the EdgeFirst Studio validation session (`v-XXXX`) that produced the measurement.
+YOLO11 Detection models trained on [COCO 2017](https://edgefirst.studio/public/projects/1684/home) (80 classes) and validated on real edge hardware through the EdgeFirst Profiler + Validator pipeline. Each row in the tables below cites the EdgeFirst Studio validation session (`v-XXXX`) that produced the measurement.
 
 Part of the [EdgeFirst Model Zoo](https://huggingface.co/spaces/EdgeFirst/Models).
 
 > [!TIP]
-> **Training experiment**: [View on EdgeFirst Studio](https://test.edgefirst.studio/public/projects/2839/experiment/training/list?exp_id=4656) — dataset, training configuration, metrics, and exported artifacts.
+> **Training experiment**: [View on EdgeFirst Studio](https://edgefirst.studio/public/projects/1684/experiment/training/list?exp_id=2115) — dataset, training configuration, metrics, and exported artifacts.
 
 > [!NOTE]
 > Architecture with C3k2 attention blocks.
@@ -63,9 +57,9 @@ Accuracy ceiling for each size, measured against COCO `val2017` (5,000 images) w
 
 | Size | Params | GFLOPs | mAP@0.5 | mAP@0.5-0.95 | mAP@0.75 | Source |
 |------|--------|--------|---------|--------------|----------|--------|
-| Nano | 2.6M | 6.5 | 53.05% | 37.82% | 41.13% | [v-1dc3](https://test.edgefirst.studio/public/validation/v-1dc3/details) |
-| Small | 9.4M | 21.5 | 61.21% | 44.91% | 48.44% | [v-1d36](https://test.edgefirst.studio/public/validation/v-1d36/details) |
-| Medium | 20.1M | 68.0 | 65.91% | 49.59% | 53.75% | [v-1dc0](https://test.edgefirst.studio/public/validation/v-1dc0/details) |
+| Nano | 2.6M | 6.5 | 53.05% | 37.78% | 40.95% | [v-51f](https://edgefirst.studio/public/validation/v-51f/details?mode=charts) |
+| Small | 9.4M | 21.5 | 61.19% | 44.89% | 48.47% | [v-520](https://edgefirst.studio/public/validation/v-520/details?mode=charts) |
+| Medium | 20.1M | 68.0 | — | — | — | — |
 | Large | 25.3M | 87.6 | — | — | — | — |
 | XLarge | 56.9M | 195.0 | — | — | — | — |
 
@@ -75,25 +69,30 @@ Accuracy ceiling for each size, measured against COCO `val2017` (5,000 images) w
 
 Each row is one EdgeFirst Studio validation session. Click the **Source** link to inspect the full session — model artifact, dataset version, parameters, per-stage Perfetto trace, and the host hardware description (hostname, kernel version, SoC, NPU, profiler version).
 
-Two row conventions in the table below:
+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 that are currently work in progress — typically a larger size that has not yet been profiled on a given NPU, or a session that has not yet been linked to an ONNX FP32 reference. The Studio Source link tracks the current status.
-- **Rows whose `Δ vs FP32` cell carries a ⚠** are published with a caveat: the quantization or runtime regression on that platform exceeds the 10 percentage-point publication threshold we grade every measurement against. The numbers are still measured on real hardware and reproducible from the linked Studio session — they are shown for transparency rather than suppressed. We are actively investigating these regressions in collaboration with the toolchain vendor and working to close the gap; the note immediately under the table summarises the current state.
-
-| Size | Platform | mAP@0.5 | Δ vs FP32 (pp) | mAP@0.5-0.95 | Inference (ms) | FPS (median) | Source |
-|------|----------|---------|----------------|--------------|----------------|--------------|--------|
-| Nano | ONNX FP32 (CPU) | 53.05% | +0.00 | 37.83% | 6.21 | 142.5 | [v-1d35](https://test.edgefirst.studio/public/validation/v-1d35/details) |
-| Nano | ONNX FP32 | 53.04% | -0.01 | 37.81% | 6.56 | 150.8 | [v-1dc4](https://test.edgefirst.studio/public/validation/v-1dc4/details) |
-| Nano | NXP i.MX 8M Plus + VeriSilicon NPU | 49.91% | -3.14 | 35.22% | 88.33 | 9.2 | [v-1d4b](https://test.edgefirst.studio/public/validation/v-1d4b/details) |
-| Nano | NXP i.MX 95 + eIQ Neutron NPU | 11.83% | -41.22 ⚠ | 6.08% | 26.45 | 31.9 | [v-1d51](https://test.edgefirst.studio/public/validation/v-1d51/details) |
-| Nano | Raspberry Pi 5 + Hailo-8L NPU | 51.28% | -1.77 | 36.37% | 18.92 | 51.1 | [v-1d3a](https://test.edgefirst.studio/public/validation/v-1d3a/details) |
-| Small | NXP i.MX 95 + eIQ Neutron NPU | 0.16% | -61.05 ⚠ | 0.09% | 53.12 | 17.4 | [v-1d55](https://test.edgefirst.studio/public/validation/v-1d55/details) |
-| Small | Raspberry Pi 5 + Hailo-8L NPU | 59.92% | -1.29 | 43.71% | 38.93 | 25.2 | [v-1d3d](https://test.edgefirst.studio/public/validation/v-1d3d/details) |
-| Medium | ONNX FP32 (CPU) | 65.91% | +0.00 | 49.59% | 9.66 | 100.3 | [v-1d37](https://test.edgefirst.studio/public/validation/v-1d37/details) |
-| Medium | ONNX FP32 | 65.86% | -0.05 | 49.48% | 19.46 | 51.0 | [v-1dc2](https://test.edgefirst.studio/public/validation/v-1dc2/details) |
-| Medium | Raspberry Pi 5 + Hailo-8L NPU | 64.56% | -1.35 | 48.30% | 72.74 | 13.6 | [v-1d40](https://test.edgefirst.studio/public/validation/v-1d40/details) |
-
-> **⚠ Investigations in progress.** The rows marked ⚠ above show a drop of more than 10 percentage points against the same training session's ONNX FP32 reference. We publish them as-is so the current state of every (model, platform) pair is visible; we are working with the toolchain vendor to close each regression, and the numbers will improve as fixes land. The next snapshot of this card will reflect any recovered accuracy.
+- **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.
+- **INT8 rows use the smart (multi-scale split) decoder** wherever both smart and logical decoder validations exist — smart is the variant we publish and ship. A dedicated section comparing the smart and logical decoders, using these benchmark results as reference cards, is planned.
+- **End-to-end (ms)** is the sequential per-image wall time covering the *full* pipeline — image load → JPEG decode → preprocess → inference → postprocess. **Throughput (FPS)** is the measured pipelined rate over the same full pipeline, which normally exceeds `1000 / end-to-end` because the runtime overlaps stages across frames.
+
+| Size | Platform | mAP@0.5 | Δ vs FP32 (pp) | mAP@0.5-0.95 | Inference (ms) | End-to-end (ms) | Throughput (FPS) | Source |
+|------|----------|---------|----------------|--------------|----------------|-----------------|------------------|--------|
+| Nano | macOS CoreML — Neural Engine (FP16) | 53.05% | +0.00 | 37.82% | 1.83 | 5.62 | 720.5 | [v-516](https://edgefirst.studio/public/validation/v-516/details?mode=charts) |
+| Nano | macOS CoreML — Metal GPU (FP16) | 53.04% | -0.01 | 37.80% | 4.26 | 8.30 | 412.5 | [v-51e](https://edgefirst.studio/public/validation/v-51e/details?mode=charts) |
+| Nano | macOS CoreML — CPU (FP16) | 53.05% | ref | 37.78% | 15.69 | 20.37 | 124.5 | [v-51f](https://edgefirst.studio/public/validation/v-51f/details?mode=charts) |
+| Nano | NXP i.MX 8M Plus + VeriSilicon NPU | 47.49% | -5.56 | 32.17% | 92.70 | 122.21 | 9.7 | [v-52a](https://edgefirst.studio/public/validation/v-52a/details?mode=charts) |
+| Nano | NXP i.MX 95 + eIQ Neutron NPU | 10.77% | -42.28 ⚠ | 5.27% | 50.94 | 84.77 | 17.7 | [v-52b](https://edgefirst.studio/public/validation/v-52b/details?mode=charts) |
+| Nano | Raspberry Pi 5 + Hailo-8L NPU | 50.90% | -2.15 | 36.02% | 17.86 | 28.43 | 53.7 | [v-4de](https://edgefirst.studio/public/validation/v-4de/details?mode=charts) |
+| Nano | NVIDIA Jetson Orin Nano (TensorRT FP16) | 53.03% | -0.02 | 37.81% | 7.09 | 17.56 | 161.4 | [v-4dc](https://edgefirst.studio/public/validation/v-4dc/details?mode=charts) |
+| Small | macOS CoreML — Neural Engine (FP16) | 61.17% | -0.02 | 44.88% | 4.95 | 8.74 | 328.9 | [v-517](https://edgefirst.studio/public/validation/v-517/details?mode=charts) |
+| Small | macOS CoreML — Metal GPU (FP16) | 61.19% | ref | 44.89% | 9.95 | 13.73 | 194.5 | [v-520](https://edgefirst.studio/public/validation/v-520/details?mode=charts) |
+| Small | macOS CoreML — CPU (FP16) | 61.13% | -0.06 | 44.86% | 31.94 | 36.54 | 60.0 | [v-521](https://edgefirst.studio/public/validation/v-521/details?mode=charts) |
+| Small | Raspberry Pi 5 + Hailo-8L NPU | 58.88% | -2.31 | 42.84% | 36.53 | 47.06 | 25.2 | [v-4df](https://edgefirst.studio/public/validation/v-4df/details?mode=charts) |
+| Small | NVIDIA Jetson Orin Nano (TensorRT FP16) | 61.13% | -0.06 | 44.85% | 15.57 | 24.81 | 109.0 | [v-4dd](https://edgefirst.studio/public/validation/v-4dd/details?mode=charts) |
+
+> **⚠ Below expectations — under investigation.** The rows marked ⚠ above measure more than 10 percentage points below the same training session's float reference: the model accuracy on that platform is below our expectations. We publish the measured numbers rather than hiding them, and we are investigating the results to make improvements — the next snapshot of this card will reflect any recovered accuracy.
 
 
 ---
@@ -102,11 +101,11 @@ Two row conventions in the table below:
 
 These results are produced by the EdgeFirst on-target validation pipeline:
 
-1. **EdgeFirst Profiler** runs on the target hardware, executes the full inference pipeline (preprocess → inference → postprocess), and emits per-image predictions in EdgeFirst Arrow/Parquet plus a Perfetto trace.
+1. **EdgeFirst Profiler** runs on the target hardware, executes the full inference pipeline (image load → decode → preprocess → inference → postprocess), and emits per-image predictions in EdgeFirst Arrow/Parquet plus a Perfetto trace.
 2. **EdgeFirst Validator** consumes the predictions and trace, computes `pycocotools` accuracy metrics and per-stage timing summaries, and publishes the results to the Studio validation session.
 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. FPS is the measured end-to-end pipelined throughput from the Perfetto trace, which generally exceeds `1000 / (preprocess + inference + postprocess)` because the runtime overlaps stages across frames.
+Inference latency is reported as the on-accelerator inference time. End-to-end latency is the sequential per-image wall time across the full pipeline — image load, JPEG decode, preprocessing, inference, and postprocessing — and throughput is the measured pipelined FPS from the Perfetto trace over that same full pipeline. Throughput generally exceeds `1000 / end-to-end` because the runtime overlaps stages across frames.
 
 See [EdgeFirst Studio](https://edgefirst.studio) for the full validation pipeline.
 
@@ -146,9 +145,11 @@ for det in results.detections:
 Every measurement in the tables above is reachable through the EdgeFirst Studio validation framework. The `v-XXXX` Source link on each row resolves to a public Studio URL of the form:
 
 ```
-https://test.edgefirst.studio/public/validation/v-XXXX/details
+https://edgefirst.studio/public/validation/v-XXXX/details?mode=charts
 ```
 
+The link lands on the **Charts** view — live system traces (CPU, memory, temperature, power) and per-stage timing recorded during the validation run. The Info and Metrics tabs on the same page carry the configuration and full COCO metric breakdown.
+
 From there, the full provenance chain is one click deeper: training session ID, dataset version, calibration artifact, converter chain (e.g. TFLite quantizer + Neutron compile), validation parameters, and the host hardware description (hostname, kernel version, SoC, NPU, profiler version). The same model file you download from this repository embeds the same chain in its `edgefirst.json` metadata.
 
 ---