Hugging Face's logo

Ultralytics
/
YOLO26

Object Detection
ultralytics
yolo
instance-segmentation
image-classification
pose-estimation
obb
tracking
semantic-segmentation
yolo26
Eval Results (legacy)
Model card Files
xet
 Community
5

update model section tasks with all metrics to match with Ultralytics github readme

#4
by onuralpszr - opened
base: refs/heads/main
←
from: refs/pr/4
Discussion Files changed
+64
-42
Files changed (1) hide show
  1. README.md +64 -42
README.md CHANGED
@@ -142,13 +142,13 @@ Ultralytics supports a wide range of YOLO models, from early versions like [YOLO
142
 
143
  Explore the [Detection Docs](https://docs.ultralytics.com/tasks/detect) for usage examples. These models are trained on the [COCO dataset](https://cocodataset.org/), featuring 80 object classes.
144
 
145
- | Model | size<br><sup>(pixels)</sup> | mAP<sup>val<br>50-95</sup> | mAP<sup>val<br>50-95(e2e)</sup> | Speed<br><sup>CPU ONNX<br>(ms)</sup> | Speed<br><sup>T4 TensorRT10<br>(ms)</sup> | params<br><sup>(M)</sup> | FLOPs<br><sup>(B)</sup> |
146
- | ------------------------------------------------------------------------------------ | --------------------------- | -------------------------- | ------------------------------- | ------------------------------------ | ----------------------------------------- | ------------------------ | ----------------------- |
147
- | [YOLO26n](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26n.pt) | 640 | 40.9 | 40.1 | 38.9 ± 0.7 | 1.7 ± 0.0 | 2.4 | 5.4 |
148
- | [YOLO26s](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26s.pt) | 640 | 48.6 | 47.8 | 87.2 ± 0.9 | 2.5 ± 0.0 | 9.5 | 20.7 |
149
- | [YOLO26m](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26m.pt) | 640 | 53.1 | 52.5 | 220.0 ± 1.4 | 4.7 ± 0.1 | 20.4 | 68.2 |
150
- | [YOLO26l](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26l.pt) | 640 | 55.0 | 54.4 | 286.2 ± 2.0 | 6.2 ± 0.2 | 24.8 | 86.4 |
151
- | [YOLO26x](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26x.pt) | 640 | 57.5 | 56.9 | 525.8 ± 4.0 | 11.8 ± 0.2 | 55.7 | 193.9 |
152
 
153
  - **mAP<sup>val</sup>** values refer to single-model single-scale performance on the [COCO val2017](https://cocodataset.org/) dataset. See [YOLO Performance Metrics](https://docs.ultralytics.com/guides/yolo-performance-metrics) for details. <br>Reproduce with `yolo val detect data=coco.yaml device=0`
154
  - **Speed** metrics are averaged over COCO val images using an [Amazon EC2 P4d](https://aws.amazon.com/ec2/instance-types/p4/) instance. CPU speeds measured with [ONNX](https://onnx.ai/) export. GPU speeds measured with [TensorRT](https://developer.nvidia.com/tensorrt) export. <br>Reproduce with `yolo val detect data=coco.yaml batch=1 device=0|cpu`
@@ -159,13 +159,13 @@ Explore the [Detection Docs](https://docs.ultralytics.com/tasks/detect) for usag
159
 
160
  Refer to the [Segmentation Docs](https://docs.ultralytics.com/tasks/segment) for usage examples. These models are trained on [COCO-Seg](https://docs.ultralytics.com/datasets/segment/coco), including 80 classes.
161
 
162
- | Model | size<br><sup>(pixels)</sup> | mAP<sup>box<br>50-95(e2e)</sup> | mAP<sup>mask<br>50-95(e2e)</sup> | Speed<br><sup>CPU ONNX<br>(ms)</sup> | Speed<br><sup>T4 TensorRT10<br>(ms)</sup> | params<br><sup>(M)</sup> | FLOPs<br><sup>(B)</sup> |
163
- | -------------------------------------------------------------------------------------------- | --------------------------- | ------------------------------- | -------------------------------- | ------------------------------------ | ----------------------------------------- | ------------------------ | ----------------------- |
164
- | [YOLO26n-seg](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26n-seg.pt) | 640 | 39.6 | 33.9 | 53.3 ± 0.5 | 2.1 ± 0.0 | 2.7 | 9.1 |
165
- | [YOLO26s-seg](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26s-seg.pt) | 640 | 47.3 | 40.0 | 118.4 ± 0.9 | 3.3 ± 0.0 | 10.4 | 34.2 |
166
- | [YOLO26m-seg](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26m-seg.pt) | 640 | 52.5 | 44.1 | 328.2 ± 2.4 | 6.7 ± 0.1 | 23.6 | 121.5 |
167
- | [YOLO26l-seg](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26l-seg.pt) | 640 | 54.4 | 45.5 | 387.0 ± 3.7 | 8.0 ± 0.1 | 28.0 | 139.8 |
168
- | [YOLO26x-seg](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26x-seg.pt) | 640 | 56.5 | 47.0 | 787.0 ± 6.8 | 16.4 ± 0.1 | 62.8 | 313.5 |
169
 
170
  - **mAP<sup>val</sup>** values are for single-model single-scale on the [COCO val2017](https://cocodataset.org/) dataset. See [YOLO Performance Metrics](https://docs.ultralytics.com/guides/yolo-performance-metrics) for details. <br>Reproduce with `yolo val segment data=coco.yaml device=0`
171
  - **Speed** metrics are averaged over COCO val images using an [Amazon EC2 P4d](https://aws.amazon.com/ec2/instance-types/p4/) instance. CPU speeds measured with [ONNX](https://onnx.ai/) export. GPU speeds measured with [TensorRT](https://developer.nvidia.com/tensorrt) export. <br>Reproduce with `yolo val segment data=coco.yaml batch=1 device=0|cpu`
@@ -176,30 +176,51 @@ Refer to the [Segmentation Docs](https://docs.ultralytics.com/tasks/segment) for
176
 
177
  See the [Semantic Segmentation Docs](https://docs.ultralytics.com/tasks/semantic) for usage examples. These models are trained on [Cityscapes](https://docs.ultralytics.com/datasets/semantic/cityscapes), including 19 classes.
178
 
179
- | Model | size<br><sup>(pixels)</sup> | mIoU<sup>val</sup> | Speed<br><sup>RTX3090 PyTorch<br>(ms)</sup> | params<br><sup>(M)</sup> | FLOPs<br><sup>(B)</sup> |
180
- | -------------------------------------------------------------------------------------------- | --------------------------- | ------------------ | ------------------------------------------- | ------------------------ | ----------------------- |
181
- | [YOLO26n-sem](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26n-sem.pt) | 1024 &times; 2048 | 78.3 | 4.4 ± 0.0 | 1.6 | 22.7 |
182
- | [YOLO26s-sem](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26s-sem.pt) | 1024 &times; 2048 | 80.8 | 8.4 ± 0.0 | 6.5 | 88.8 |
183
- | [YOLO26m-sem](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26m-sem.pt) | 1024 &times; 2048 | 82.0 | 19.9 ± 0.1 | 14.3 | 304.5 |
184
- | [YOLO26l-sem](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26l-sem.pt) | 1024 &times; 2048 | 82.9 | 26.5 ± 0.1 | 17.9 | 384.7 |
185
- | [YOLO26x-sem](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26x-sem.pt) | 1024 &times; 2048 | 83.6 | 48.9 ± 0.2 | 40.2 | 861.7 |
186
 
187
  - **mIoU<sup>val</sup>** values are for single-model single-scale on the [Cityscapes](https://www.cityscapes-dataset.com/) validation set. <br>Reproduce with `yolo semantic val data=cityscapes.yaml device=0 imgsz=2048`
188
  - **Speed** metrics are averaged over Cityscapes validation images using an RTX3090 instance. <br>Reproduce with `yolo semantic val data=cityscapes.yaml batch=1 device=0|cpu imgsz=2048`
189
 
190
  </details>
191
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
192
  <details><summary>Classification (ImageNet)</summary>
193
 
194
  Consult the [Classification Docs](https://docs.ultralytics.com/tasks/classify) for usage examples. These models are trained on [ImageNet](https://docs.ultralytics.com/datasets/classify/imagenet), covering 1000 classes.
195
 
196
- | Model | size<br><sup>(pixels)</sup> | acc<br><sup>top1</sup> | acc<br><sup>top5</sup> | Speed<br><sup>CPU ONNX<br>(ms)</sup> | Speed<br><sup>T4 TensorRT10<br>(ms)</sup> | params<br><sup>(M)</sup> | FLOPs<br><sup>(B) at 224</sup> |
197
- | -------------------------------------------------------------------------------------------- | --------------------------- | ---------------------- | ---------------------- | ------------------------------------ | ----------------------------------------- | ------------------------ | ------------------------------ |
198
- | [YOLO26n-cls](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26n-cls.pt) | 224 | 71.4 | 90.1 | 5.0 ± 0.3 | 1.1 ± 0.0 | 2.8 | 0.5 |
199
- | [YOLO26s-cls](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26s-cls.pt) | 224 | 76.0 | 92.9 | 7.9 ± 0.2 | 1.3 ± 0.0 | 6.7 | 1.6 |
200
- | [YOLO26m-cls](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26m-cls.pt) | 224 | 78.1 | 94.2 | 17.2 ± 0.4 | 2.0 ± 0.0 | 11.6 | 4.9 |
201
- | [YOLO26l-cls](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26l-cls.pt) | 224 | 79.0 | 94.6 | 23.2 ± 0.3 | 2.8 ± 0.0 | 14.1 | 6.2 |
202
- | [YOLO26x-cls](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26x-cls.pt) | 224 | 79.9 | 95.0 | 41.4 ± 0.9 | 3.8 ± 0.0 | 29.6 | 13.6 |
203
 
204
  - **acc** values represent model accuracy on the [ImageNet](https://www.image-net.org/) dataset validation set. <br>Reproduce with `yolo val classify data=path/to/ImageNet device=0`
205
  - **Speed** metrics are averaged over ImageNet val images using an [Amazon EC2 P4d](https://aws.amazon.com/ec2/instance-types/p4/) instance. CPU speeds measured with [ONNX](https://onnx.ai/) export. GPU speeds measured with [TensorRT](https://developer.nvidia.com/tensorrt) export. <br>Reproduce with `yolo val classify data=path/to/ImageNet batch=1 device=0|cpu`
@@ -210,13 +231,13 @@ Consult the [Classification Docs](https://docs.ultralytics.com/tasks/classify) f
210
 
211
  See the [Pose Estimation Docs](https://docs.ultralytics.com/tasks/pose) for usage examples. These models are trained on [COCO-Pose](https://docs.ultralytics.com/datasets/pose/coco), focusing on the 'person' class.
212
 
213
- | Model | size<br><sup>(pixels)</sup> | mAP<sup>pose<br>50-95(e2e)</sup> | mAP<sup>pose<br>50(e2e)</sup> | Speed<br><sup>CPU ONNX<br>(ms)</sup> | Speed<br><sup>T4 TensorRT10<br>(ms)</sup> | params<br><sup>(M)</sup> | FLOPs<br><sup>(B)</sup> |
214
- | ---------------------------------------------------------------------------------------------- | --------------------------- | -------------------------------- | ----------------------------- | ------------------------------------ | ----------------------------------------- | ------------------------ | ----------------------- |
215
- | [YOLO26n-pose](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26n-pose.pt) | 640 | 57.2 | 83.3 | 40.3 ± 0.5 | 1.8 ± 0.0 | 2.9 | 7.5 |
216
- | [YOLO26s-pose](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26s-pose.pt) | 640 | 63.0 | 86.6 | 85.3 ± 0.9 | 2.7 ± 0.0 | 10.4 | 23.9 |
217
- | [YOLO26m-pose](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26m-pose.pt) | 640 | 68.8 | 89.6 | 218.0 ± 1.5 | 5.0 ± 0.1 | 21.5 | 73.1 |
218
- | [YOLO26l-pose](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26l-pose.pt) | 640 | 70.4 | 90.5 | 275.4 ± 2.4 | 6.5 ± 0.1 | 25.9 | 91.3 |
219
- | [YOLO26x-pose](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26x-pose.pt) | 640 | 71.6 | 91.6 | 565.4 ± 3.0 | 12.2 ± 0.2 | 57.6 | 201.7 |
220
 
221
  - **mAP<sup>val</sup>** values are for single-model single-scale on the [COCO Keypoints val2017](https://docs.ultralytics.com/datasets/pose/coco) dataset. See [YOLO Performance Metrics](https://docs.ultralytics.com/guides/yolo-performance-metrics) for details. <br>Reproduce with `yolo val pose data=coco-pose.yaml device=0`
222
  - **Speed** metrics are averaged over COCO val images using an [Amazon EC2 P4d](https://aws.amazon.com/ec2/instance-types/p4/) instance. CPU speeds measured with [ONNX](https://onnx.ai/) export. GPU speeds measured with [TensorRT](https://developer.nvidia.com/tensorrt) export. <br>Reproduce with `yolo val pose data=coco-pose.yaml batch=1 device=0|cpu`
@@ -227,19 +248,20 @@ See the [Pose Estimation Docs](https://docs.ultralytics.com/tasks/pose) for usag
227
 
228
  Check the [OBB Docs](https://docs.ultralytics.com/tasks/obb) for usage examples. These models are trained on [DOTAv1](https://docs.ultralytics.com/datasets/obb/dota-v2#dota-v10), including 15 classes.
229
 
230
- | Model | size<br><sup>(pixels)</sup> | mAP<sup>test<br>50-95(e2e)</sup> | mAP<sup>test<br>50(e2e)</sup> | Speed<br><sup>CPU ONNX<br>(ms)</sup> | Speed<br><sup>T4 TensorRT10<br>(ms)</sup> | params<br><sup>(M)</sup> | FLOPs<br><sup>(B)</sup> |
231
- | -------------------------------------------------------------------------------------------- | --------------------------- | -------------------------------- | ----------------------------- | ------------------------------------ | ----------------------------------------- | ------------------------ | ----------------------- |
232
- | [YOLO26n-obb](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26n-obb.pt) | 1024 | 52.4 | 78.9 | 97.7 ± 0.9 | 2.8 ± 0.0 | 2.5 | 14.0 |
233
- | [YOLO26s-obb](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26s-obb.pt) | 1024 | 54.8 | 80.9 | 218.0 ± 1.4 | 4.9 ± 0.1 | 9.8 | 55.1 |
234
- | [YOLO26m-obb](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26m-obb.pt) | 1024 | 55.3 | 81.0 | 579.2 ± 3.8 | 10.2 ± 0.3 | 21.2 | 183.3 |
235
- | [YOLO26l-obb](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26l-obb.pt) | 1024 | 56.2 | 81.6 | 735.6 ± 3.1 | 13.0 ± 0.2 | 25.6 | 230.0 |
236
- | [YOLO26x-obb](https://github.com/ultralytics/assets/releases/download/v8.4.0/yolo26x-obb.pt) | 1024 | 56.7 | 81.7 | 1485.7 ± 11.5 | 30.5 ± 0.9 | 57.6 | 516.5 |
237
 
238
  - **mAP<sup>test</sup>** values are for single-model multiscale performance on the [DOTAv1 test set](https://captain-whu.github.io/DOTA/dataset.html). <br>Reproduce by `yolo val obb data=DOTAv1.yaml device=0 split=test` and submit merged results to the [DOTA evaluation server](https://captain-whu.github.io/DOTA/evaluation.html).
239
  - **Speed** metrics are averaged over [DOTAv1 val images](https://docs.ultralytics.com/datasets/obb/dota-v2#dota-v10) using an [Amazon EC2 P4d](https://aws.amazon.com/ec2/instance-types/p4/) instance. CPU speeds measured with [ONNX](https://onnx.ai/) export. GPU speeds measured with [TensorRT](https://developer.nvidia.com/tensorrt) export. <br>Reproduce by `yolo val obb data=DOTAv1.yaml batch=1 device=0|cpu`
240
 
241
  </details>
242
 
 
243
  ## 🧩 Integrations
244
 
245
  Our key integrations with leading AI platforms extend the functionality of Ultralytics' offerings, enhancing tasks like dataset labeling, training, visualization, and model management. Discover how Ultralytics, in collaboration with partners like [Weights & Biases](https://docs.ultralytics.com/integrations/weights-biases), [Comet ML](https://docs.ultralytics.com/integrations/comet), [Roboflow](https://docs.ultralytics.com/integrations/roboflow), and [Intel OpenVINO](https://docs.ultralytics.com/integrations/openvino), can optimize your AI workflow. Explore more at [Ultralytics Integrations](https://docs.ultralytics.com/integrations).
 
142
 
143
  Explore the [Detection Docs](https://docs.ultralytics.com/tasks/detect) for usage examples. These models are trained on the [COCO dataset](https://cocodataset.org/), featuring 80 object classes.
144
 
145
+ | Model | size<br><sup>(pixels)</sup> | mAP<sup>val<br>50-95</sup> | mAP<sup>val<br>50-95(e2e)</sup> | Speed<br><sup>CPU ONNX<br>(ms)</sup> | Speed<br><sup>T4 TensorRT10<br>(ms)</sup> | params<br><sup>(M)</sup> | FLOPs<br><sup>(B)</sup> |
146
+ | ---------------------------------------------------------------------- | --------------------------- | -------------------------- | ------------------------------- | ------------------------------------ | ----------------------------------------- | ------------------------ | ----------------------- |
147
+ | [YOLO26n](https://platform.ultralytics.com/ultralytics/yolo26/yolo26n) | 640 | 40.9 | 40.1 | 38.9 ± 0.7 | 1.7 ± 0.0 | 2.4 | 5.5 |
148
+ | [YOLO26s](https://platform.ultralytics.com/ultralytics/yolo26/yolo26s) | 640 | 48.6 | 47.8 | 87.2 ± 0.9 | 2.5 ± 0.0 | 9.5 | 20.9 |
149
+ | [YOLO26m](https://platform.ultralytics.com/ultralytics/yolo26/yolo26m) | 640 | 53.1 | 52.5 | 220.0 ± 1.4 | 4.7 ± 0.1 | 20.4 | 68.4 |
150
+ | [YOLO26l](https://platform.ultralytics.com/ultralytics/yolo26/yolo26l) | 640 | 55.0 | 54.4 | 286.2 ± 2.0 | 6.2 ± 0.2 | 24.8 | 86.8 |
151
+ | [YOLO26x](https://platform.ultralytics.com/ultralytics/yolo26/yolo26x) | 640 | 57.5 | 56.9 | 525.8 ± 4.0 | 11.8 ± 0.2 | 55.7 | 194.4 |
152
 
153
  - **mAP<sup>val</sup>** values refer to single-model single-scale performance on the [COCO val2017](https://cocodataset.org/) dataset. See [YOLO Performance Metrics](https://docs.ultralytics.com/guides/yolo-performance-metrics) for details. <br>Reproduce with `yolo val detect data=coco.yaml device=0`
154
  - **Speed** metrics are averaged over COCO val images using an [Amazon EC2 P4d](https://aws.amazon.com/ec2/instance-types/p4/) instance. CPU speeds measured with [ONNX](https://onnx.ai/) export. GPU speeds measured with [TensorRT](https://developer.nvidia.com/tensorrt) export. <br>Reproduce with `yolo val detect data=coco.yaml batch=1 device=0|cpu`
 
159
 
160
  Refer to the [Segmentation Docs](https://docs.ultralytics.com/tasks/segment) for usage examples. These models are trained on [COCO-Seg](https://docs.ultralytics.com/datasets/segment/coco), including 80 classes.
161
 
162
+ | Model | size<br><sup>(pixels)</sup> | mAP<sup>box<br>50-95(e2e)</sup> | mAP<sup>mask<br>50-95(e2e)</sup> | Speed<br><sup>CPU ONNX<br>(ms)</sup> | Speed<br><sup>T4 TensorRT10<br>(ms)</sup> | params<br><sup>(M)</sup> | FLOPs<br><sup>(B)</sup> |
163
+ | ------------------------------------------------------------------------------ | --------------------------- | ------------------------------- | -------------------------------- | ------------------------------------ | ----------------------------------------- | ------------------------ | ----------------------- |
164
+ | [YOLO26n-seg](https://platform.ultralytics.com/ultralytics/yolo26/yolo26n-seg) | 640 | 39.6 | 33.9 | 53.3 ± 0.5 | 2.1 ± 0.0 | 2.7 | 9.3 |
165
+ | [YOLO26s-seg](https://platform.ultralytics.com/ultralytics/yolo26/yolo26s-seg) | 640 | 47.3 | 40.0 | 118.4 ± 0.9 | 3.3 ± 0.0 | 10.4 | 34.5 |
166
+ | [YOLO26m-seg](https://platform.ultralytics.com/ultralytics/yolo26/yolo26m-seg) | 640 | 52.5 | 44.1 | 328.2 ± 2.4 | 6.7 ± 0.1 | 23.6 | 121.7 |
167
+ | [YOLO26l-seg](https://platform.ultralytics.com/ultralytics/yolo26/yolo26l-seg) | 640 | 54.4 | 45.5 | 387.0 ± 3.7 | 8.0 ± 0.1 | 28.0 | 140.1 |
168
+ | [YOLO26x-seg](https://platform.ultralytics.com/ultralytics/yolo26/yolo26x-seg) | 640 | 56.5 | 47.0 | 787.0 ± 6.8 | 16.4 ± 0.1 | 62.8 | 314.0 |
169
 
170
  - **mAP<sup>val</sup>** values are for single-model single-scale on the [COCO val2017](https://cocodataset.org/) dataset. See [YOLO Performance Metrics](https://docs.ultralytics.com/guides/yolo-performance-metrics) for details. <br>Reproduce with `yolo val segment data=coco.yaml device=0`
171
  - **Speed** metrics are averaged over COCO val images using an [Amazon EC2 P4d](https://aws.amazon.com/ec2/instance-types/p4/) instance. CPU speeds measured with [ONNX](https://onnx.ai/) export. GPU speeds measured with [TensorRT](https://developer.nvidia.com/tensorrt) export. <br>Reproduce with `yolo val segment data=coco.yaml batch=1 device=0|cpu`
 
176
 
177
  See the [Semantic Segmentation Docs](https://docs.ultralytics.com/tasks/semantic) for usage examples. These models are trained on [Cityscapes](https://docs.ultralytics.com/datasets/semantic/cityscapes), including 19 classes.
178
 
179
+ | Model | size<br><sup>(pixels)</sup> | mIoU<sup>val</sup> | Speed<br><sup>RTX3090 PyTorch<br>(ms)</sup> | params<br><sup>(M)</sup> | FLOPs<br><sup>(B)</sup> |
180
+ | ------------------------------------------------------------------------------ | --------------------------- | ------------------ | ------------------------------------------- | ------------------------ | ----------------------- |
181
+ | [YOLO26n-sem](https://platform.ultralytics.com/ultralytics/yolo26/yolo26n-sem) | 1024 &times; 2048 | 78.3 | 4.4 ± 0.0 | 1.6 | 23.8 |
182
+ | [YOLO26s-sem](https://platform.ultralytics.com/ultralytics/yolo26/yolo26s-sem) | 1024 &times; 2048 | 80.8 | 8.4 ± 0.0 | 6.5 | 91.0 |
183
+ | [YOLO26m-sem](https://platform.ultralytics.com/ultralytics/yolo26/yolo26m-sem) | 1024 &times; 2048 | 82.0 | 19.9 ± 0.1 | 14.3 | 305.5 |
184
+ | [YOLO26l-sem](https://platform.ultralytics.com/ultralytics/yolo26/yolo26l-sem) | 1024 &times; 2048 | 82.9 | 26.5 ± 0.1 | 17.8 | 388.2 |
185
+ | [YOLO26x-sem](https://platform.ultralytics.com/ultralytics/yolo26/yolo26x-sem) | 1024 &times; 2048 | 83.6 | 48.9 ± 0.2 | 40.1 | 866.9 |
186
 
187
  - **mIoU<sup>val</sup>** values are for single-model single-scale on the [Cityscapes](https://www.cityscapes-dataset.com/) validation set. <br>Reproduce with `yolo semantic val data=cityscapes.yaml device=0 imgsz=2048`
188
  - **Speed** metrics are averaged over Cityscapes validation images using an RTX3090 instance. <br>Reproduce with `yolo semantic val data=cityscapes.yaml batch=1 device=0|cpu imgsz=2048`
189
 
190
  </details>
191
 
192
+ <details><summary>Depth Estimation (NYU Depth V2)</summary>
193
+
194
+ See the [Depth Estimation Docs](https://docs.ultralytics.com/tasks/depth) for usage examples. These models are pretrained on a broad multi-dataset mix and evaluated on the [NYU Depth V2](https://cs.nyu.edu/~fergus/datasets/nyu_depth_v2.html) Eigen test split, predicting per-pixel depth in meters.
195
+
196
+ | Model | size<br><sup>(pixels)</sup> | delta1<sup>NYU</sup> | abs_rel<sup>NYU</sup> | rmse<sup>NYU</sup> | Speed<br><sup>CPU ONNX<br>(ms)</sup> | Speed<br><sup>T4 TensorRT10<br>(ms)</sup> | params<br><sup>(M)</sup> | FLOPs<br><sup>(B)</sup> |
197
+ | ---------------------------------------------------------------------------------- | --------------------------- | -------------------- | --------------------- | ------------------ | ------------------------------------ | ----------------------------------------- | ------------------------ | ----------------------- |
198
+ | [YOLO26n-depth](https://platform.ultralytics.com/ultralytics/yolo26/yolo26n-depth) | 768 | 0.882 | 0.109 | 0.414 | 272.0 ± 27.2 | 2.7 ± 0.1 | 6.3 | 46.9 |
199
+ | [YOLO26s-depth](https://platform.ultralytics.com/ultralytics/yolo26/yolo26s-depth) | 768 | 0.896 | 0.104 | 0.399 | 393.7 ± 13.1 | 3.8 ± 0.0 | 13.2 | 68.0 |
200
+ | [YOLO26m-depth](https://platform.ultralytics.com/ultralytics/yolo26/yolo26m-depth) | 768 | 0.921 | 0.089 | 0.364 | 621.5 ± 49.7 | 6.0 ± 0.1 | 23.3 | 130.4 |
201
+ | [YOLO26l-depth](https://platform.ultralytics.com/ultralytics/yolo26/yolo26l-depth) | 768 | 0.930 | 0.083 | 0.351 | 821.9 ± 50.7 | 7.7 ± 0.1 | 27.7 | 157.0 |
202
+ | [YOLO26x-depth](https://platform.ultralytics.com/ultralytics/yolo26/yolo26x-depth) | 768 | 0.933 | 0.080 | 0.344 | 1240.9 ± 73.3 | 13.6 ± 0.2 | 57.0 | 301.7 |
203
+
204
+ - **delta1<sup>NYU</sup>** is the percentage of pixels where the predicted depth is within a factor of 1.25 of the ground truth, on the NYU Depth V2 Eigen test split (654 images) with multi-scale + horizontal-flip TTA and log-least-squares alignment.
205
+ - Single-scale accuracy without TTA is reproducible with `yolo depth val model=yolo26n-depth.pt data=nyu-depth.yaml imgsz=768 device=0` (substitute `model=` for each size), which uses median (scale-only) alignment and scores lower: delta1 0.785 (n), 0.786 (s), 0.827 (m), 0.839 (l), 0.843 (x).
206
+ - **abs_rel** is the mean absolute relative error between predicted and ground-truth depth values.
207
+ - **rmse** is the root mean squared error in meters.
208
+ - **Speed** is inference-only latency (pre/post-processing excluded) at `imgsz=768`, `batch=1`, reported as mean ± std over timed runs after warmup. **CPU ONNX** is ONNX Runtime fp32 on a 32-core Intel Xeon (Skylake); **T4 TensorRT10** is TensorRT fp16 on a Tesla T4.
209
+ - **params** and **FLOPs** are measured at 768×768, the training resolution of the released weights.
210
+
211
+ </details>
212
+
213
  <details><summary>Classification (ImageNet)</summary>
214
 
215
  Consult the [Classification Docs](https://docs.ultralytics.com/tasks/classify) for usage examples. These models are trained on [ImageNet](https://docs.ultralytics.com/datasets/classify/imagenet), covering 1000 classes.
216
 
217
+ | Model | size<br><sup>(pixels)</sup> | acc<br><sup>top1</sup> | acc<br><sup>top5</sup> | Speed<br><sup>CPU ONNX<br>(ms)</sup> | Speed<br><sup>T4 TensorRT10<br>(ms)</sup> | params<br><sup>(M)</sup> | FLOPs<br><sup>(B) at 224</sup> |
218
+ | ------------------------------------------------------------------------------ | --------------------------- | ---------------------- | ---------------------- | ------------------------------------ | ----------------------------------------- | ------------------------ | ------------------------------ |
219
+ | [YOLO26n-cls](https://platform.ultralytics.com/ultralytics/yolo26/yolo26n-cls) | 224 | 71.4 | 90.1 | 5.0 ± 0.3 | 1.1 ± 0.0 | 2.8 | 0.4 |
220
+ | [YOLO26s-cls](https://platform.ultralytics.com/ultralytics/yolo26/yolo26s-cls) | 224 | 76.0 | 92.9 | 7.9 ± 0.2 | 1.3 ± 0.0 | 6.7 | 1.5 |
221
+ | [YOLO26m-cls](https://platform.ultralytics.com/ultralytics/yolo26/yolo26m-cls) | 224 | 78.1 | 94.2 | 17.2 ± 0.4 | 2.0 ± 0.0 | 11.6 | 4.8 |
222
+ | [YOLO26l-cls](https://platform.ultralytics.com/ultralytics/yolo26/yolo26l-cls) | 224 | 79.0 | 94.6 | 23.2 ± 0.3 | 2.8 ± 0.0 | 14.1 | 6.0 |
223
+ | [YOLO26x-cls](https://platform.ultralytics.com/ultralytics/yolo26/yolo26x-cls) | 224 | 79.9 | 95.0 | 41.4 ± 0.9 | 3.8 ± 0.0 | 29.6 | 13.5 |
224
 
225
  - **acc** values represent model accuracy on the [ImageNet](https://www.image-net.org/) dataset validation set. <br>Reproduce with `yolo val classify data=path/to/ImageNet device=0`
226
  - **Speed** metrics are averaged over ImageNet val images using an [Amazon EC2 P4d](https://aws.amazon.com/ec2/instance-types/p4/) instance. CPU speeds measured with [ONNX](https://onnx.ai/) export. GPU speeds measured with [TensorRT](https://developer.nvidia.com/tensorrt) export. <br>Reproduce with `yolo val classify data=path/to/ImageNet batch=1 device=0|cpu`
 
231
 
232
  See the [Pose Estimation Docs](https://docs.ultralytics.com/tasks/pose) for usage examples. These models are trained on [COCO-Pose](https://docs.ultralytics.com/datasets/pose/coco), focusing on the 'person' class.
233
 
234
+ | Model | size<br><sup>(pixels)</sup> | mAP<sup>pose<br>50-95(e2e)</sup> | mAP<sup>pose<br>50(e2e)</sup> | Speed<br><sup>CPU ONNX<br>(ms)</sup> | Speed<br><sup>T4 TensorRT10<br>(ms)</sup> | params<br><sup>(M)</sup> | FLOPs<br><sup>(B)</sup> |
235
+ | -------------------------------------------------------------------------------- | --------------------------- | -------------------------------- | ----------------------------- | ------------------------------------ | ----------------------------------------- | ------------------------ | ----------------------- |
236
+ | [YOLO26n-pose](https://platform.ultralytics.com/ultralytics/yolo26/yolo26n-pose) | 640 | 57.2 | 83.3 | 40.3 ± 0.5 | 1.8 ± 0.0 | 2.9 | 7.6 |
237
+ | [YOLO26s-pose](https://platform.ultralytics.com/ultralytics/yolo26/yolo26s-pose) | 640 | 63.0 | 86.6 | 85.3 ± 0.9 | 2.7 ± 0.0 | 10.4 | 24.1 |
238
+ | [YOLO26m-pose](https://platform.ultralytics.com/ultralytics/yolo26/yolo26m-pose) | 640 | 68.8 | 89.6 | 218.0 ± 1.5 | 5.0 ± 0.1 | 21.5 | 73.3 |
239
+ | [YOLO26l-pose](https://platform.ultralytics.com/ultralytics/yolo26/yolo26l-pose) | 640 | 70.4 | 90.5 | 275.4 ± 2.4 | 6.5 ± 0.1 | 25.9 | 91.7 |
240
+ | [YOLO26x-pose](https://platform.ultralytics.com/ultralytics/yolo26/yolo26x-pose) | 640 | 71.6 | 91.6 | 565.4 ± 3.0 | 12.2 ± 0.2 | 57.6 | 202.3 |
241
 
242
  - **mAP<sup>val</sup>** values are for single-model single-scale on the [COCO Keypoints val2017](https://docs.ultralytics.com/datasets/pose/coco) dataset. See [YOLO Performance Metrics](https://docs.ultralytics.com/guides/yolo-performance-metrics) for details. <br>Reproduce with `yolo val pose data=coco-pose.yaml device=0`
243
  - **Speed** metrics are averaged over COCO val images using an [Amazon EC2 P4d](https://aws.amazon.com/ec2/instance-types/p4/) instance. CPU speeds measured with [ONNX](https://onnx.ai/) export. GPU speeds measured with [TensorRT](https://developer.nvidia.com/tensorrt) export. <br>Reproduce with `yolo val pose data=coco-pose.yaml batch=1 device=0|cpu`
 
248
 
249
  Check the [OBB Docs](https://docs.ultralytics.com/tasks/obb) for usage examples. These models are trained on [DOTAv1](https://docs.ultralytics.com/datasets/obb/dota-v2#dota-v10), including 15 classes.
250
 
251
+ | Model | size<br><sup>(pixels)</sup> | mAP<sup>test<br>50-95(e2e)</sup> | mAP<sup>test<br>50(e2e)</sup> | Speed<br><sup>CPU ONNX<br>(ms)</sup> | Speed<br><sup>T4 TensorRT10<br>(ms)</sup> | params<br><sup>(M)</sup> | FLOPs<br><sup>(B)</sup> |
252
+ | ------------------------------------------------------------------------------ | --------------------------- | -------------------------------- | ----------------------------- | ------------------------------------ | ----------------------------------------- | ------------------------ | ----------------------- |
253
+ | [YOLO26n-obb](https://platform.ultralytics.com/ultralytics/yolo26/yolo26n-obb) | 1024 | 52.4 | 78.9 | 97.7 ± 0.9 | 2.8 ± 0.0 | 2.4 | 14.8 |
254
+ | [YOLO26s-obb](https://platform.ultralytics.com/ultralytics/yolo26/yolo26s-obb) | 1024 | 54.8 | 80.9 | 218.0 ± 1.4 | 4.9 ± 0.1 | 9.8 | 56.7 |
255
+ | [YOLO26m-obb](https://platform.ultralytics.com/ultralytics/yolo26/yolo26m-obb) | 1024 | 55.3 | 81.0 | 579.2 ± 3.8 | 10.2 ± 0.3 | 21.2 | 184.9 |
256
+ | [YOLO26l-obb](https://platform.ultralytics.com/ultralytics/yolo26/yolo26l-obb) | 1024 | 56.2 | 81.6 | 735.6 ± 3.1 | 13.0 ± 0.2 | 25.6 | 232.4 |
257
+ | [YOLO26x-obb](https://platform.ultralytics.com/ultralytics/yolo26/yolo26x-obb) | 1024 | 56.7 | 81.7 | 1485.7 ± 11.5 | 30.5 ± 0.9 | 57.6 | 520.1 |
258
 
259
  - **mAP<sup>test</sup>** values are for single-model multiscale performance on the [DOTAv1 test set](https://captain-whu.github.io/DOTA/dataset.html). <br>Reproduce by `yolo val obb data=DOTAv1.yaml device=0 split=test` and submit merged results to the [DOTA evaluation server](https://captain-whu.github.io/DOTA/evaluation.html).
260
  - **Speed** metrics are averaged over [DOTAv1 val images](https://docs.ultralytics.com/datasets/obb/dota-v2#dota-v10) using an [Amazon EC2 P4d](https://aws.amazon.com/ec2/instance-types/p4/) instance. CPU speeds measured with [ONNX](https://onnx.ai/) export. GPU speeds measured with [TensorRT](https://developer.nvidia.com/tensorrt) export. <br>Reproduce by `yolo val obb data=DOTAv1.yaml batch=1 device=0|cpu`
261
 
262
  </details>
263
 
264
+
265
  ## 🧩 Integrations
266
 
267
  Our key integrations with leading AI platforms extend the functionality of Ultralytics' offerings, enhancing tasks like dataset labeling, training, visualization, and model management. Discover how Ultralytics, in collaboration with partners like [Weights & Biases](https://docs.ultralytics.com/integrations/weights-biases), [Comet ML](https://docs.ultralytics.com/integrations/comet), [Roboflow](https://docs.ultralytics.com/integrations/roboflow), and [Intel OpenVINO](https://docs.ultralytics.com/integrations/openvino), can optimize your AI workflow. Explore more at [Ultralytics Integrations](https://docs.ultralytics.com/integrations).