What I did today
trained 3 tf hub models on my data but only 2000 images of 20000,found that resnet one works the best
still the mse was aroung 150 for age prediction which is quite bad,but i believe this is due to using only 2000 images so should improve by a good amount on 20000
didnt change the gender pred code that much,took the same resnet_50 model i was using in last version
What I'll do tomorrow
my company training bootcamp focusing java starts so thats 9 to 5 , after that ill preactice and apply those things
was thinking that i could keep my model on training on the whole set during when my training sessions are going on,lets see
i want to research ways of improving tf hu models ,ztm has a page on it,so will explore that
5 coding questions from leetcode