596236 (1) [Avatar] Offline
#1
One of the steps mentions to train the custom network/model (step 3). How come the code does not show that. Shouldn't the custom layers be compiled and trained before adding it to the vgg model and then trained again like mentioned in step 5. I am having difficult time following the code since it does not reflect what is mentioned in the theory.
I have similar issues with feature extractions and ensemble.
Can someone please correct me if if am wrong and just following the code as in the example will lead us to the best possible accuracy.
Thank You,
Kal