I need directions on how to do transfer learning with Yolov3 in pytorch

I have about 400 images all labeled with correct anchor boxes from supervisely and I want to apply object detection on them. I am trying to understand the exact steps I need to get everything working? My current thought process is to first find out where I can grab darknet from pytorch like VGG and just apply transfer learning with my dataset. I can probably just change the input shape and the output vector because they are the only things that are going to be different. Can someone please help me and point me to the right direction please?

I would search for good implementations of object detection models on github.
This YOLO3 version seems to provide a training script.
Also @chenyuntc has provided a nice repo for Faster RCNN here.