How to transfer learn Darknet YOLOv3

I've started getting into object detection in image. I have YOLOv3 neural network with Darknet framework. The network is pre-trained from COCO data set. Now I need to do some transfer learning in order to try to make the results better. What I have so far:

I've created my custom data set and did transfer learning following this guide: https://github.com/ultralytics/yolov3/wiki/Example:-Transfer-Learning

After few hours it spit out some .pt file. From what I gathered .pt is PyTorch format but my program uses .weights which is Darknet format. First I've tried to find how to convert .pt to .weights and I stumbled upon this: https://github.com/marvis/pytorch-caffe-darknet-convert but apparently this works only with YOLOv2.

Then I tried to look for something on how to transfer learn using Darknet but with no luck.

Since then I'm stuck and can't move forward, can any of you give me some tips and pointers on how to achieve transfer learning on YOLOLv3 + Darknet (or how to convert .pt to .weights since I already know how to create .pt)? Thanks!

• Hi and welcome to this community! I think that this question is more suited for Data Science SE, given that it involves some implementation issues. Ask it there ;) – nbro Sep 25 at 15:56
• Let me know if you'd like me to migrate it to Data Science for you. – DukeZhou Sep 25 at 22:52