LA Tran

Results 4 issues of LA Tran

In Tensorflow 1.x version, we use multi-scale training, why we discard it in Tensorflow 2.x? Is that an issue of using Keras?

After training the model, I would like to get the output from each branch separately, but I found that only small-object branch has detection outputs and it includes large objects...

The model works well until I want to train large-object branch only! In training phase, I freeze medium-object branch and small-object branch (set "trainable=False"), and I only calculate loss using...

Hi, Thanks for sharing. What a nice work! I have not been able to figure out why does CDDC block yield 53 channels? Thanks, LA Tran