Training Accuracy cannot go above 99%. How can I improve it?

H​i everybody,

I​ am having problems reaching the required accuracy and I am not sure what lever to try.

I​ am using binary cross-entropy as a loss function and RMSProp or Adam algorithm for the training. Also, I am using 3 convolutions for image compression/reduction.

I​ played with different values learning rates, and different numbers of neurons in the hidden layer but still the model cannot exceed 90% training accuracy.

[code removed - moderator]

Thanks,
Thang

Please click my name and message your notebook as an attachment.

[code removed - moderator]

After trying with 3 Conv layers. I test it on 4 Conv layers in my notebook and got a new model.
I attachment model.summary(). Thank you mentor.

This assignment wants you to push the training accuracy to atleast 99.9%. Why are you augmenting images with transformations other than rescaling?

My model is working again. Thank you.