This demo showcases top-down pipeline for human pose estimation on video or image. The task is to predict bboxes for every person on frame and then to predict a pose for every detected person. The pose may contain up to 17 keypoints: ears, eyes, nose, shoulders, elbows, wrists, hips, knees, and ankles.
On the start-up, the application reads command line parameters and loads detection person model and single human pose estimation model. Upon getting a frame from the OpenCV VideoCapture, the demo executes top-down pipeline for this frame and displays the results.
Running the application with the
-h option yields the following usage message:
To run the demo, you can use public or pre-trained models. To download the pre-trained models, use the OpenVINO Model Downloader. The list of models supported by the demo is in the
models.lst file in the demo's directory.
For example, to do inference on a CPU, run the following command:
The demo uses OpenCV to display the resulting frame with estimated poses and reports performance in the following format: summary inference FPS (single human pose inference FPS / detector inference FPS).