Flask & Keras project for forest mushrooms classification.
In this project I used the MyShroomClassifier Image Dataset. I've removed all monochrome images from this dataset and also I've balanced out the classes using augmentation. After such manipulations each class of mushrooms was consisting of ~2500 images. Keras framework was used in order to train two neural networks:
- CNN (~70% accuracy on test dataset)
- Transfer learning network, VGG19 (~80% accuracy on test dataset)
To provide the ability for user to pass the images to CNNs and to receive predictions from them Flask web-framework was used.
