Use timm pretrained backbones for you FasterRCNN!
-
model_config.py -> it returns the model,feat_sizes,output channel and the feat layer names, which is reqd by the Add_FPN.py file
-
Add_FPN.py -> Edited the BackboneWithFPN function from pytorch, which is now used to add FPN to any timm model, till now it can be only used for Efficentnet family
-
test.ipynb -> an example explaining how to use it any backbone and add FPN to it
Loading even small models takes up huge amout of memory, I cant quite detect why, although the training and inferencing speed is maintained. (For example if we add effnet_b1 in the FasterRcnn, which is smaller in terms of params of Resnet101,which is a default option in FasterRcnn still it occupies larger space in memory)