Applying pose estimation to videos

[caleb-vicente]

Which model is better?

• Movenet
• https://github.com/MVIG-SJTU/AlphaPose. To use AlphaPose folow this tutotial (https://github.com/Amanbhandula/AlphaPose)

[caleb-vicente]

Alphapose is already working in Google Colab, the main problem is that it doesn´t detect players perfectly well, most likely because two reasons:

• Model has been pretrained with people not with American football players
• It is using yolo v3 instead of v7.

As future recommendation, we sugest to use yolo v7 object detection already implemented and then apply movenet for each bounding box indiviudally