mnist | learn mnist classifier with very high accuracy | Machine Learning library

I assigned mnist as:

I know it's basic and too easy for you people, but I'm a beginner who needs your help. I'm struggling to make binary classifier with CNN. My final goal is to check accuracy over 0.99

I import both MNIST and FASHION_MNIST to identify if it's number or clothing. So there are 2 category. I want to categorize 0-60000 as 0, and 60001-120000 as 1. I will use binary_crossentropy.

but I dont know how to start from the beginning. How can I use vstack hstack at first to combine MNIST and FASHION_MNIST?

This is how I tried so far

I'm trying to build an LSTM encoder. I'm testing it on the MNIST dataset to check any errors before using it on my actual dataset. My code:

I have a set of images of numbers that go from 0 to 20 with intermediate classes(0,25 / 0,5 / 0,75). Each number will be defined as a class of its own. I have 22 images per class.

These images will be used for training and testing on a convolutional neural network for classification. I'm not worried about accuracy, it's only a proof of concept so I realise that the dataset is too small for any real reliable outcome. Like I said, it's only meant as a proof of concept.

EDIT: As suggested by @Kaveh, I checked out ImageDataGenerator.flow_from_directory

As far as I could tell, this is used to increase your dataset size using data augmentation. However, what I'm asking is, now that I have these images set in different folders (22 images per folder, each folder making a class) how do I use them. I've always been loading one file that makes up the dataset (example: mnist; through keras). I've never used my own data and therefore, have no idea what the next step is.

I am implementing an autoencoder using the Fashion Mnsit dataset. The code for the encoder-

I'm doing keras subclassing with the mnist dataset. I was able to make it with Sequantial and Functional api's. But now when i call model.fit() on my subclass i get this error:

I am trying to implement a simple GAN in google collaboratory, After using transforms to normalize the images, I want to view it at the output end to display fake image generated by the generator and real image side by in the dataset once every batch iteration like a video.

Please find the below TF Keras Model in which I am using tanh activation function in the Hidden Layers.

While the value of Logits are proper, the values that are calculated by implementing the tanh function manually is resulting in Nan.

It may be because of the Runtime Warnings shown below:

/home/abc/anaconda3/lib/python3.7/site-packages/ipykernel_launcher.py:76: RuntimeWarning: overflow encountered in exp

/home/abc/anaconda3/lib/python3.7/site-packages/ipykernel_launcher.py:76: RuntimeWarning: invalid value encountered in true_divide

Complete reproducible code is mentioned below:

• I'm have studied about Autoencoder and tried to implement a simple one.
• I have built a model with one hidden layer.
• I Run it with mnist digits dataset and plot the digits before the Autoencoder and after it.
• I saw some examples which used hidden layer of size 32 or 64, I tried it and it didn't gave the same (or something close to) the source images.
• I tried to change the hidden layer to size of 784 (same as the input size, just to test the model) but got same results.

What am I missing ? Why the examples on the web shows good results and when I test it, I'm getting different results ?