Which part should be frozen during transfer learning?

Question

I want to use transfer learning and fine tuning and I need to decide which part of
the original model will be used and which part will be frozen. I'm thinking about four possilbe cases:
small/large new dataset and this set is similar/not similar to the original dataset. What should be done to achieve best results in each of the cases?

Saandeep Sreerambatla · Accepted Answer

Ideally I follow the below rules for performing transfer learning, I could do successful object detection using Yolo model by traning only last layer and was able to do leaf classification by training good number of layers.

Which part should be frozen during transfer learning?

One Answer

Add your own answers!

Ask a Question