Is there any expected loss in training using SSD? Say I train one class detection and use 1200 images in training. What would be my expected loss when training is optimum? We normally stop training when loss doesn't change much? How can I verify that loss is optimum for my training?
Asked
Active
Viewed 57 times
0
-
Optimum loss is 0. Whether your net can theoretically reach this optimum one cannot say. Whether it is a good thing to reach optimum = 0 error on the training set is a wholly different question. – Jan Kukacka Jun 29 '18 at 10:28
-
@JanKukacka hi thanks. Say my training achieved minimum loss 5. What are the things I need to look at in training. – batuman Jun 30 '18 at 06:06
-
@JanKukacka, I need to come back with some plots for my training so that you have some ideas. I'll come back with plots for loss and learning rate. – batuman Jun 30 '18 at 06:29