Open zengyiwen123 opened 7 months ago
Our demo code can be divided into 3 phases. First we use a detection model to detect players in the video (here), and then we select the largest & highest confident bounding bounding box (here). Finally, we clip the image using the selected bbox and run HybrIK (here).
Therefore, to support multiplayer scenes, you only need to change phase 2 and select more than 1 bound bboxes.
When running the demo, the multiplayer scene can only perform pose estimation for a single person, can hybrik implement the pose estimation in the multiplayer scenario?