Digit classification with tensorflow: https://d-costa.github.io/digits/
- Optional centering: understand what areas are more relevant to identify digits;
- Probability for each digit;
The MNIST dataset is used for training a CNN network with the following architecture:
| Layer | Output Shape | Param # |
|---|---|---|
| InputLayer | [(None, 28, 28, 1)] | 0 |
| Conv2D (32, (3, 3), relu) | (None, 26, 26, 32) | 320 |
| MaxPooling2D (2,2) | (None, 13, 13, 32) | 0 |
| Conv2D (16, (3, 3), relu) | (None, 11, 11, 16) | 4624 |
| MaxPooling2 | (None, 5, 5, 16) | 0 |
| Flatten | (None, 400) | 0 |
| Dense (relu) | (None, 128) | 51328 |
| Dense (softmax) | (None, 10) | 1290 |
- Total params: 57,562
- Trainable params: 57,562
- Non-trainable params: 0
- ~15.000 examples of the training set are used as validation
- Adam is used as the optimizer with 15 epochs and batch size of 64.
Test loss: 0.050
Test accuracy: 98.97%
-
Install the required python packages:
pip install -r requirements.txt
-
Modify the model architecture in train.py and train the model:
python train.py
The model is saved in a file named model.h5.
-
Test the trained model:
python test.py
-
Convert the model to use it in tensorflowjs:
./convert.sh
The model is saved in /public/tfjs
-
Open index.html in your browser
- Carlos Aguayo : https://github.com/carlos-aguayo/carlos-aguayo.github.io
- CNN architecture render: https://netron.app/