[Docs] add the support information about MMPose Hourglass (#1118)

* [Docs] add the support information about MMPose Hourglass * * add Hourglass support config * update benchmark doc * modify Hourglass AP under pytorch backend * add regression test for Hourglass * *update 03-benchmark/benchmark.md * * modify mmpose.yml to ignore ncnn int8 test * add trt_fp16 test results * * add openvino test result * * modify supported_models.md
2025-01-14 08:09:43 +08:00 · 2022-10-10 16:38:54 +08:00 · 2022-10-10 16:38:54 +08:00 · fd21b98efa
commit fd21b98efa
parent 41fcc2242f
10 changed files with 90 additions and 2 deletions
--- a/configs/mmpose/pose-detection_ncnn_static-256x256.py
+++ b/configs/mmpose/pose-detection_ncnn_static-256x256.py
@ -0,0 +1,3 @@
+_base_ = ['./pose-detection_static.py', '../_base_/backends/ncnn.py']
+
+onnx_config = dict(input_shape=[256, 256])
--- a/configs/mmpose/pose-detection_openvino_static-256x256.py
+++ b/configs/mmpose/pose-detection_openvino_static-256x256.py
@ -0,0 +1,5 @@
+_base_ = ['./pose-detection_static.py', '../_base_/backends/openvino.py']
+
+onnx_config = dict(input_shape=[256, 256])
+backend_config = dict(
+    model_inputs=[dict(opt_shapes=dict(input=[1, 3, 256, 256]))])
--- a/configs/mmpose/pose-detection_tensorrt_static-256x256.py
+++ b/configs/mmpose/pose-detection_tensorrt_static-256x256.py
@ -0,0 +1,13 @@
+_base_ = ['./pose-detection_static.py', '../_base_/backends/tensorrt.py']
+
+onnx_config = dict(input_shape=[256, 256])
+backend_config = dict(
+    common_config=dict(max_workspace_size=1 << 30),
+    model_inputs=[
+        dict(
+            input_shapes=dict(
+                input=dict(
+                    min_shape=[1, 3, 256, 256],
+                    opt_shape=[1, 3, 256, 256],
+                    max_shape=[1, 3, 256, 256])))
+    ])
--- a/docs/en/03-benchmark/benchmark.md
+++ b/docs/en/03-benchmark/benchmark.md
@ -1689,6 +1689,27 @@ Users can directly test the performance through [how_to_evaluate_a_model.md](../
    <td align="center">-</td>
    <td align="center">0.825</td>
  </tr>
+  <tr>
+    <td align="center" rowspan="2"><a href="https://github.com/open-mmlab/mmpose/blob/master/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hourglass52_coco_256x256.py">Hourglass</a></td>
+    <td align="center" rowspan="2">Pose Detection</td>
+    <td align="center" rowspan="2">COCO</td>
+    <td align="center">AP</td>
+    <td align="center">0.717</td>
+    <td align="center">0.717</td>
+    <td align="center">0.717</td>
+    <td align="center">0.717</td>
+    <td align="center">-</td>
+    <td align="center">0.717</td>
+  </tr>
+  <tr>
+    <td align="center">AR</td>
+    <td align="center">0.774</td>
+    <td align="center">0.774</td>
+    <td align="center">0.774</td>
+    <td align="center">0.774</td>
+    <td align="center">-</td>
+    <td align="center">0.774</td>
+  </tr>
 </tbody>
 </table>
 </div>
--- a/docs/en/03-benchmark/quantization.md
+++ b/docs/en/03-benchmark/quantization.md
@ -31,6 +31,6 @@ Note:  [mmocr](https://github.com/open-mmlab/mmocr)  Uses 'shapely' to compute I

 |                                                                      model                                                                       | dataset  | fp32 AP | int8 AP |
 | :----------------------------------------------------------------------------------------------------------------------------------------------: | :------: | :-----: | :-----: |
-| [Hourglass](https://github.com/open-mmlab/mmpose/blob/master/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hourglass52_coco_256x256.py) | COCO2017 |  0.726  |  0.713  |
+| [Hourglass](https://github.com/open-mmlab/mmpose/blob/master/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hourglass52_coco_256x256.py) | COCO2017 |  0.717  |  0.713  |

 Note: MMPose models are tested with `flip_test` explicitly set to `False` in model configs.
--- a/docs/en/03-benchmark/supported_models.md
+++ b/docs/en/03-benchmark/supported_models.md
@ -73,6 +73,7 @@ The table below lists the models that are guaranteed to be exportable to other b
 | HRNet                       | MMPose           |      N      |      Y      |    Y     |  Y   |   N   |    Y     |   N    |  N   |     [config](https://mmpose.readthedocs.io/en/latest/papers/backbones.html#hrnet-cvpr-2019)     |
 | MSPN                        | MMPose           |      N      |      Y      |    Y     |  Y   |   N   |    Y     |   N    |  N   |     [config](https://mmpose.readthedocs.io/en/latest/papers/backbones.html#mspn-arxiv-2019)     |
 | LiteHRNet                   | MMPose           |      N      |      Y      |    Y     |  N   |   N   |    Y     |   N    |  N   |   [config](https://mmpose.readthedocs.io/en/latest/papers/backbones.html#litehrnet-cvpr-2021)   |
+| Hourglass                   | MMPose           |      N      |      Y      |    Y     |  Y   |   N   |    Y     |   N    |  N   |   [config](https://mmpose.readthedocs.io/en/latest/papers/backbones.html#hourglass-eccv-2016)   |
 | PointPillars                | MMDetection3d    |      ?      |      Y      |    Y     |  N   |   N   |    Y     |   N    |  N   |     [config](https://github.com/open-mmlab/mmdetection3d/blob/master/configs/pointpillars)      |
 | CenterPoint (pillar)        | MMDetection3d    |      ?      |      Y      |    Y     |  N   |   N   |    Y     |   N    |  N   |      [config](https://github.com/open-mmlab/mmdetection3d/blob/master/configs/centerpoint)      |
 | RotatedRetinaNet            | RotatedDetection |      N      |      Y      |    Y     |  N   |   N   |    N     |   N    |  N   | [config](https://github.com/open-mmlab/mmrotate/blob/main/configs/rotated_retinanet/README.md)  |
--- a/docs/zh_cn/03-benchmark/benchmark.md
+++ b/docs/zh_cn/03-benchmark/benchmark.md
@ -1665,6 +1665,27 @@ GPU: ncnn, TensorRT, PPLNN
    <td align="center">-</td>
    <td align="center">0.825</td>
  </tr>
+  <tr>
+    <td align="center" rowspan="2"><a href="https://github.com/open-mmlab/mmpose/blob/master/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hourglass52_coco_256x256.py">Hourglass</a></td>
+    <td align="center" rowspan="2">Pose Detection</td>
+    <td align="center" rowspan="2">COCO</td>
+    <td align="center">AP</td>
+    <td align="center">0.717</td>
+    <td align="center">0.717</td>
+    <td align="center">0.717</td>
+    <td align="center">0.717</td>
+    <td align="center">-</td>
+    <td align="center">0.717</td>
+  </tr>
+  <tr>
+    <td align="center">AR</td>
+    <td align="center">0.774</td>
+    <td align="center">0.774</td>
+    <td align="center">0.774</td>
+    <td align="center">0.774</td>
+    <td align="center">-</td>
+    <td align="center">0.774</td>
+  </tr>
 </tbody>
 </table>
 </div>
--- a/docs/zh_cn/03-benchmark/quantization.md
+++ b/docs/zh_cn/03-benchmark/quantization.md
@ -31,6 +31,6 @@

 |                                                                      model                                                                       | dataset  | fp32 AP | int8 AP |
 | :----------------------------------------------------------------------------------------------------------------------------------------------: | :------: | :-----: | :-----: |
-| [Hourglass](https://github.com/open-mmlab/mmpose/blob/master/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hourglass52_coco_256x256.py) | COCO2017 |  0.726  |  0.713  |
+| [Hourglass](https://github.com/open-mmlab/mmpose/blob/master/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hourglass52_coco_256x256.py) | COCO2017 |  0.717  |  0.713  |

 备注：测试转换后的模型精度时，对于 mmpose 模型，在模型配置文件中 `flip_test` 需设置为 `False`。
--- a/docs/zh_cn/03-benchmark/supported_models.md
+++ b/docs/zh_cn/03-benchmark/supported_models.md
@ -73,6 +73,7 @@
 | HRNet                       | MMPose           |      N      |      Y      |    Y     |  Y   |   N   |    Y     |   N    |  N   |     [config](https://mmpose.readthedocs.io/en/latest/papers/backbones.html#hrnet-cvpr-2019)     |
 | MSPN                        | MMPose           |      N      |      Y      |    Y     |  Y   |   N   |    Y     |   N    |  N   |     [config](https://mmpose.readthedocs.io/en/latest/papers/backbones.html#mspn-arxiv-2019)     |
 | LiteHRNet                   | MMPose           |      N      |      Y      |    Y     |  N   |   N   |    Y     |   N    |  N   |   [config](https://mmpose.readthedocs.io/en/latest/papers/backbones.html#litehrnet-cvpr-2021)   |
+| Hourglass                   | MMPose           |      N      |      Y      |    Y     |  Y   |   N   |    Y     |   N    |  N   |   [config](https://mmpose.readthedocs.io/en/latest/papers/backbones.html#hourglass-eccv-2016)   |
 | PointPillars                | MMDetection3d    |      ?      |      Y      |    Y     |  N   |   N   |    Y     |   N    |  N   |     [config](https://github.com/open-mmlab/mmdetection3d/blob/master/configs/pointpillars)      |
 | CenterPoint (pillar)        | MMDetection3d    |      ?      |      Y      |    Y     |  N   |   N   |    Y     |   N    |  N   |      [config](https://github.com/open-mmlab/mmdetection3d/blob/master/configs/centerpoint)      |
 | RotatedRetinaNet            | RotatedDetection |      N      |      Y      |    Y     |  N   |   N   |    N     |   N    |  N   | [config](https://github.com/open-mmlab/mmrotate/blob/main/configs/rotated_retinanet/README.md)  |
--- a/tests/regression/mmpose.yml
+++ b/tests/regression/mmpose.yml
@ -36,6 +36,11 @@ tensorrt:
    backend_test: *default_backend_test
    sdk_config: *sdk_static
    deploy_config: configs/mmpose/pose-detection_tensorrt_static-256x192.py
+  pipeline_trt_static_fp32_256x256: &pipeline_trt_static_fp32_256x256
+    convert_image: *convert_image
+    backend_test: *default_backend_test
+    sdk_config: *sdk_static
+    deploy_config: configs/mmpose/pose-detection_tensorrt_static-256x256.py
  pipeline_trt_static_fp16: &pipeline_trt_static_fp16
    convert_image: *convert_image
    backend_test: *default_backend_test
@ -52,12 +57,20 @@ openvino:
    convert_image: *convert_image
    backend_test: *default_backend_test
    deploy_config: configs/mmpose/pose-detection_openvino_static-256x192.py
+  pipeline_openvino_static_fp32_256x256: &pipeline_openvino_static_fp32_256x256
+    convert_image: *convert_image
+    backend_test: *default_backend_test
+    deploy_config: configs/mmpose/pose-detection_openvino_static-256x256.py

 ncnn:
  pipeline_ncnn_static_fp32: &pipeline_ncnn_static_fp32
    convert_image: *convert_image
    backend_test: False
    deploy_config: configs/mmpose/pose-detection_ncnn_static-256x192.py
+  pipeline_ncnn_static_fp32_256x256: &pipeline_ncnn_static_fp32_256x256
+    convert_image: *convert_image
+    backend_test: False
+    deploy_config: configs/mmpose/pose-detection_ncnn_static-256x256.py

 pplnn:
  pipeline_pplnn_static_fp32: &pipeline_pplnn_static_fp32
@ -100,3 +113,13 @@ models:
      - *pipeline_trt_static_fp16
      - *pipeline_ncnn_static_fp32
      - *pipeline_openvino_static_fp32
+
+  - name: Hourglass
+    metafile: configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hourglass_coco.yml
+    model_configs:
+      - configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hourglass52_coco_256x256.py
+    pipelines:
+      - *pipeline_ort_static_fp32
+      - *pipeline_trt_static_fp32_256x256
+      - *pipeline_ncnn_static_fp32_256x256
+      - *pipeline_openvino_static_fp32_256x256