I want to use Yolov4 and Yolov3, however the pre-trained model from darknet trained on ImageNet provided by AlexejAB Repo on Github has classes with dogs, cats,...
My small dataset is about distinguishing cables which consists of 4 classes (240 pictures in total).
Since I do not think that there is a pre-trained model with similar dataset out there, should I train from scratch or is there an advantage of transfer learning / fine tuning a large pre-trained model in my case?
I think you should first simply fine-tune the existing model on your dataset because training from scratch would be a little expensive step.