I want to do fine-tuning based on Inception V3 pre-trained on imagenet and while I am aware that the number of epochs is a hyperparameter, I would like to know what is a rule of thumb for selecting a starter number of epochs?
number of 512x512 jpeg images in each subfolder based on label:
-train
–label_0: 134,889
–label_1: 32,882
-val
–label_0: 42,553
–label_1: 8,799
-test
–label_0: 45,230
–label_1: 11,051
Should I start with 20 epochs or is it very low? Any suggestion is really appreciated.