Tool for Pose estimation from .mp4 video. Based on Movenet model (https://tfhub.dev/google/movenet/singlepose/thunder/4)
- Docker (Docker Desktop or Docker Engine (https://docs.docker.com/desktop/))
- Git (https://github.com/git-guides/install-git)
To install sc-intregnet
run in terminal:
git clone https://github.com/Barabaika/Pose_estimation_tool.git
cd Pose_estimation_tool && ./install.sh
Go to Pose_estimation_tool folder, and then:
./run_pose_estimation.sh
Go to Pose_estimation_tool folder, and then:
./run_pose_estimation.sh <abs_path_to_input_video.mp4>*
*- you can add multiple paths
Results will be stored at Pose_estimation_tool/Pose_estim_result/
From each input video it would be generated:
- _<input_video_name>output.mp4 (video with keypoints and edges)
- _<input_video_name>output.json ({'poses':{'pose_1':list_keypoints, ..., 'pose_n':list_keypoints}})