princeton-vl/CornerNet-Lite

Freezing model to train small dataset

Opened this issue · 1 comments

Hi, thanks for the code.
I was able to train now, But since my dataset is small and not very general. I want to freeze part of the model to keep the general extraction. I used yours params from your pretrained cornernet-squeeze. Can you tell me which part of the model should I freeze?

@heilaw can you help me out? Thank you