RunningLeon 787ebc2392
[Feature]: Support mmpose (#94)
* add mmpose code

* update

* update

* add rewrites

* test trt

* test litehrnet with trt

* revert unused change

* add docs about mmpose

* add docstring and staticmethod

* update

* update

* update docs

* fix config name and docs

* add pose_detection ut

* add pose data

* fix lint of model.py

* add pose_detection_model ut

* fix docs and docstrinf

* add test_mmpose_models.py

* fix yapf

* fix lint

* fix create input

* support ort ut

* fix yapf

* fix docs

* fix createinput

* test ci bug

* rm test1.py

* fix yapf

* fix flake8

* fix yapf

* add config and update benchmark

* fix table format

* update mmpose benchmark

* update benchmark for mmpose

* run mmpose tests seperately in ci

* fix lint

* resolve comments

* add trt ut config

* fix test

* fix tests

* resolve comments

* resolve comments

* update tests

Co-authored-by: VVsssssk <shenkun@pjlab.org.cn>
Co-authored-by: hanrui1sensetime <hanrui1@sensetime.com>
2022-02-16 11:03:12 +08:00

1.8 KiB

MMPose Support

MMPose is an open-source toolbox for pose estimation based on PyTorch. It is a part of the OpenMMLab project.

MMPose installation tutorial

Please refer to official installation guide to install the codebase.

MMEditing models support

Model Task ONNX Runtime TensorRT NCNN PPLNN OpenVINO Model config
HRNet PoseDetection Y Y Y N Y config
MSPN PoseDetection Y Y Y N Y config
LiteHRNet PoseDetection Y Y N N Y config

Example

python tools/deploy.py \
configs/mmpose/posedetection_tensorrt_static-256x192.py \
$MMPOSE_DIR/configs/body/2d_kpt_sview_rgb_img/topdown_heatmap/coco/hrnet_w48_coco_256x192.py \
$MMPOSE_DIR/checkpoints/hrnet_w48_coco_256x192-b9e0b3ab_20200708.pth \
$MMDEPLOY_DIR/demo/resources/human-pose.jpg \
--work-dir work-dirs/mmpose/topdown/hrnet/trt \
--device cuda

Note

  • Usually, mmpose models need some extra information for the input image, but we can't get it directly. So, when exporting the model, you can use $MMDEPLOY_DIR/demo/resources/human-pose.jpg as input.

Reminder

None

FAQs

None