Implement the LeNet-5 deep neural network model.
- Load the MNIST data
- Preprocess Data
- Setup TensorFlow
Layer | Description |
---|---|
Input | 32x32x1 grayscale image |
Convolution 5x5 | 1x1 stride, same padding, outputs 28x28x6 |
RELU | |
Max pooling | 2x2 stride, outputs 14x14x6 |
Convolution 5x5 | 1x1 stride, same padding, outputs 10x10x16 |
RELU | |
Max pooling | 2x2 stride, outputs 5x5x16 = 400 |
Fully connected | outputs 120 |
RELU | |
Dropout | Optional |
Fully connected | outputs 84 |
RELU | |
Dropout | Optional |
Fully connected | outputs 10 |
- Training and evaluation