One-hot -> Num (28281)

What it do?

This model just convert one-hot vector (shapes 1,10) in picture 28x28x1 px

Model`s structure:

1. Input: shape = (1,10)
2. Dense: 12544, activation='relu'
3. BatchNormalization
4. Reshape: (7, 7, 256)
5. Conv2DTranspose: units=128 kernel=5 strides=(1, 1) padding='same' activation='relu'
6. BatchNormalization
7. Conv2DTranspose: units=64 kernel=5 strides=(2, 2) padding='same' activation='relu'
8. BatchNormalization
9. Conv2DTranspose: units=1 kernel=5 strides=(2, 2) padding='same' activation='sigmoid'

I take model from it

If u want to use it:

1. Method 1:

import keras

model = keras.models.load_model('/path/')
model.compile(loss='mae', optimizer='adam', metrics=['acc'])

...

2. Method 2

go to there and start for modules