Hi @isarandi
Can you explain in layman's terms how to synthesize data for training to obtain 2.5D heatmaps or robust heatmaps, how to obtain the third axis for 2D images, and how the loss function is calculated?

1. According to my understanding, does it use the contour of the joint relative to the size of the root to distinguish the distance of the Z-axis?
2. Can heat maps be visualized?
3. Can reducing the number of joints accelerate reasoning?
   thanks