search

GeekAlexis / FastMOT

High-performance multiple object tracking based on YOLO, Deep SORT, and KLT 🚀
MIT License
1.15k stars 253 forks source link

Input shape of the crowdhuman model #229

Closed jmudy closed 2 years ago

jmudy commented 2 years ago

I would like to know what is the input shape of the crowdhuman model that was trained for FastMOT. Is it 512x512 or 608x608?

Greetings in advance.

GeekAlexis commented 2 years ago

It was trained with 608x608 and later reshaped into 512x512 after converting to ONNX