Model selection for small objects

I’m training a 2-class descriptor in TLT. One is human, the other is card. But card objects are very small. Which model would you recommend?

We recommend you to raise this query in TLT forum for better assistance.


You can try yolo_v4 network.