Update MODEL_ZOO.md

mjlbach · Nov 8, 2022 · 555071e · 555071e
1 parent ec6e12d
commit 555071e
Showing 1 changed file with 21 additions and 15 deletions.
diff --git a/docs/MODEL_ZOO.md b/docs/MODEL_ZOO.md
@@ -35,6 +35,27 @@ python scripts/demo_inference.py --cfg configs/halpe_26/resnet/256x192_res50_lr1
 - This model is trained based on the first 26 keypoints of Halpe Full-body datatset (without face and hand keypoints).
 - The speed is tested on COCO val2017 on a single NVIDIA GeForce RTX 3090 gpu, with `batch_size=64` in each iteration and offline yolov3 human detection results.
 
+
+## Multi Domain Models **(Strongly Recommended)** 
+| Model                    | Backbone | Detector | Input Size | Loss Type |     AP     | Speed |  Download | Config | #keypoints |
+|--------------------------|----------|----------|------------|------------|------------|-------|-----------|--------|--------------|
+|[Fast Pose](../configs/halpe_coco_wholebody_136/resnet/256x192_res50_lr1e-3_2x-regression.yaml)    | ResNet50           | YOLOv3 | 256x192            | Symmetric Integral | 50.1       | 16.28 iter/s | [Google](https://drive.google.com/file/d/1Bb3kPoFFt-M0Y3ceqNO8DTXi1iNDd4gI/view?usp=sharing) [Baidu(code: d0wi)](https://pan.baidu.com/s/1GaHzMHTqYze2rVn7u1sjVg) | [cfg](../configs/halpe_coco_wholebody_136/resnet/256x192_res50_lr1e-3_2x-regression.yaml)    | 136 |
+|[Fast Pose (DCN)](../configs/halpe_coco_wholebody_136/resnet/256x192_res50_lr1e-3_2x-dcn-combined.yaml)    | ResNet50 - dcn           | YOLOv3 | 256x192            | Combined (10 hand weight) | 49.8        | 10.35 iter/s | [Google](https://drive.google.com/file/d/1wX1Z2ZOoysgSNovlgiEtJKpbR8tUBWYR/view?usp=sharing) [Baidu(code: app1)](https://pan.baidu.com/s/1bIro0XfYj0FIVf84QzdDoQ) | [cfg](../configs/halpe_coco_wholebody_136/resnet/256x192_res50_lr1e-3_2x-dcn-combined.yaml)    | 136 |
+|[Fast Pose (DCN)](../configs/halpe_68_noface/resnet/256x192_res50_lr1e-3_2x-dcn-combined.yaml)    | ResNet50 - dcn           | YOLOv3 | 256x192            | Combined | -        | 13.88 iter/s | [Google](https://drive.google.com/file/d/14Qn9gxm-EVzqFi7v25Y5TqKIvrFLy_BR/view?usp=sharing) [Baidu(code: 6kwr)](https://pan.baidu.com/s/1GLNxN3gfekUVY0HZu41fJQ) | [cfg](../configs/halpe_68_noface/resnet/256x192_res50_lr1e-3_2x-dcn-combined.yaml)    | 68 (no face) |
+|[Fast Pose (DCN)](../configs/single_hand/resnet/256x192_res50_lr1e-3_2x-dcn-regression.yaml)    | ResNet50 - dcn           | - | 256x192            | Symmetric Integral | -        | 30.20 iter/s | [Google](https://drive.google.com/file/d/1MntndimlUP5Hxef1UN9ZDMBVglfA606J/view?usp=sharing) [Baidu(code: nwxx)](https://pan.baidu.com/s/1OR-uH25MFQ7kY8Gt_aJfbw ) | [cfg](../configs/single_hand/resnet/256x192_res50_lr1e-3_2x-dcn-regression.yaml)    | 21 (single hand) |
+
+For the most accurate wholebody pose estimation, you can run with:
+```
+python scripts/demo_inference.py --cfg configs/halpe_coco_wholebody_136/resnet/256x192_res50_lr1e-3_2x-dcn-combined.yaml --checkpoint pretrained_models/multi_domain_fast50_dcn_combined_256x192.pth --indir examples/demo/ --save_img
+```
+or, you can run with (this version is a little faster and more accurate on body keypoints, but its performance on hand keypoints is worser):
+```
+python scripts/demo_inference.py --cfg configs/halpe_coco_wholebody_136/resnet/256x192_res50_lr1e-3_2x-regression.yaml --checkpoint pretrained_models/multi_domain_fast50_regression_256x192.pth --indir examples/demo/ --save_img
+```
+#### Notes
+- Above models are trained on multiple datasets, thus they can perform well for in-the-wild images.
+- 'Combined (10 hand weight)' means that we use different loss for hand and body keypoints
+
 ## [Halpe dataset](https://github.com/Fang-Haoshu/Halpe-FullBody) (136 keypoints)
 
 | Model                    | Backbone | Detector | Input Size | Loss Type |     AP     | Speed |  Download | Config |
@@ -73,22 +94,7 @@ python scripts/demo_inference.py --cfg configs/halpe_136/resnet/256x192_res50_lr
 - The APs are tested under COCO WholeBody's criterion, with flip test on.
 - The speed is tested on COCO val2017 on a single NVIDIA GeForce RTX 3090 gpu, with `batch_size=64` in each iteration and offline yolov3 human detection results.
 
-## Multi Domain Models **(Strongly Recommended)**
-| Model                    | Backbone | Detector | Input Size | Loss Type |     AP     | Speed |  Download | Config | #keypoints |
-|--------------------------|----------|----------|------------|------------|------------|-------|-----------|--------|--------------|
-|[Fast Pose](../configs/halpe_coco_wholebody_136/resnet/256x192_res50_lr1e-3_2x-regression.yaml)    | ResNet50           | YOLOv3 | 256x192            | Symmetric Integral | 50.1       | 16.28 iter/s | [Google](https://drive.google.com/file/d/1Bb3kPoFFt-M0Y3ceqNO8DTXi1iNDd4gI/view?usp=sharing) [Baidu(code: d0wi)](https://pan.baidu.com/s/1GaHzMHTqYze2rVn7u1sjVg) | [cfg](../configs/halpe_136/resnet/256x192_res50_lr1e-3_2x-regression.yaml)    | 136 |
-|[Fast Pose (DCN)](../configs/halpe_136/resnet/256x192_res50_lr1e-3_2x-dcn-combined.yaml)    | ResNet50 - dcn           | YOLOv3 | 256x192            | Combined (10 hand weight) | 49.8        | 10.35 iter/s | [Google](https://drive.google.com/file/d/1wX1Z2ZOoysgSNovlgiEtJKpbR8tUBWYR/view?usp=sharing) [Baidu(code: app1)](https://pan.baidu.com/s/1bIro0XfYj0FIVf84QzdDoQ) | [cfg](../configs/halpe_136/resnet/256x192_res50_lr1e-3_2x-dcn-combined.yaml)    | 136 |
-|[Fast Pose (DCN)](../configs/halpe_68_noface/resnet/256x192_res50_lr1e-3_2x-dcn-combined.yaml)    | ResNet50 - dcn           | YOLOv3 | 256x192            | Combined | -        | 13.88 iter/s | [Google](https://drive.google.com/file/d/14Qn9gxm-EVzqFi7v25Y5TqKIvrFLy_BR/view?usp=sharing) [Baidu(code: 6kwr)](https://pan.baidu.com/s/1GLNxN3gfekUVY0HZu41fJQ) | [cfg](../configs/halpe_68_noface/resnet/256x192_res50_lr1e-3_2x-dcn-combined.yaml)    | 68 (no face) |
-|[Fast Pose (DCN)](../configs/single_hand/resnet/256x192_res50_lr1e-3_2x-dcn-regression.yaml)    | ResNet50 - dcn           | - | 256x192            | Symmetric Integral | -        | 30.20 iter/s | [Google](https://drive.google.com/file/d/1MntndimlUP5Hxef1UN9ZDMBVglfA606J/view?usp=sharing) [Baidu(code: nwxx)](https://pan.baidu.com/s/1OR-uH25MFQ7kY8Gt_aJfbw ) | [cfg](../configs/single_hand/resnet/256x192_res50_lr1e-3_2x-dcn-regression.yaml)    | 21 (single hand) |
 
-For the most accurate wholebody pose estimation, you can run with:
-```
-python scripts/demo_inference.py --cfg configs/halpe_coco_wholebody_136/resnet/256x192_res50_lr1e-3_2x-dcn-combined.yaml --checkpoint pretrained_models/multi_domain_fast50_dcn_combined_256x192.pth --indir examples/demo/ --save_img
-```
-or, you can run with (this version is a little faster and more accurate on body keypoints, but its performance on hand keypoints is worser):
-```
-python scripts/demo_inference.py --cfg configs/halpe_coco_wholebody_136/resnet/256x192_res50_lr1e-3_2x-regression.yaml --checkpoint pretrained_models/multi_domain_fast50_regression_256x192.pth --indir examples/demo/ --save_img
-```
 
 #### Notes
 - These models are strongly recommended because they are more accurate and flexible.