Transfer Learning experiments on Kaggle's "Mushrooms classification - Common genus's images" data set using ResNet and EfficientNet.
The data set can be found here (download the data as it is and double check that all images are inside a folder called "Mushrooms").
fitv2.py: Start training the model (either ResNet or EfficientNet) on the training data set. At first, the fully-connected layer we add on top of the pretrained network is trained for a small number of epoches. Then, in the fine-tuning phase, the training process is resumed and applied also to the pretrained convnet's topmost convolutional layer;model.py: File containing functions that return ready-to-use instances of ResNet and EfficientNet;evalmodel.py: Evaluate the trained model on a handful of images (a checkpoint is given as an example).