CNN model that coded from scratch using numpy that simply recognize the two mnist classes: 0s and 1s.
To do CNN model, I implement my mini-model-constructor using classes for each layers:
- Convolutional layer
- Linear layer
- ReLU layer
- Softmax layer
Model architecture: Conv -> Conv -> Linear -> ReLU -> Linear -> Softmax
In the end, I check same pytorch model to be sure that my implementation is correct