Accuracy is very poor

#1

i used mobilenet architecture to solve the pokedex problem but the accuracy is crossing even 15 % i tried finetuning and data augumentation but it doesnt increase the accuracy

#2

**************** not crossing

#3

Hey @dibakarchaudhary58, make sure you use 1. train test split, 2. normalizing images by dividing it by 255.0. 3. and formost important, you use the labels same as given in the question.

Hope this resolved your doubt.
Plz mark the doubt as resolved in my doubts section. :blush:

#4

is it neccessary to normalize as the mobilenet has batch normalisation layers???

#5

Hey @dibakarchaudhary58, yes i normalized it and recieved 60% validation accuracy in first epoch only.