Same as Pixel Food Classificator, but, unlike it, was done with Keras module of Tensorflow.
- Two convolutional layers with 3x3 kernel and 64 filters and ReLU activation function
- Dense layer with 64 neurons and ReLU
- Dense layer with 3 neurons and softmax activation
Some plots from Tensorboard:
Convert images to input format of shape (Nimages, 32, 32, 1):
python prepare_data.py
Train model:
python train_model.py
Test model:
python test_model.py
Check tensorboard:
tensorboard --logdir=logs/