This tool lets you build an image classifier with a single terminal command by using resnet50 as a feature extractor.

This tool requires fast.ai and Python3 to be installed. Instructions to install fast.ai can be found here
Clone the repo and navigate to the folder on Terminal.

Data must be arranged into 3 folders, train test and valid, in the form:
/data
    /train
        /class1
        /class2
        /class3
    /valid
        /class1
        /class2
        /class3
    /test

This can be done by installing the split-folders library

• Path: Compulsory parameter that specifies path to the folder containing training data.
• batchsize: Sets the batchsize(defaults of 4)
• epochs: Sets the number of epochs the model will train(defaults to 10)

Run the main.py script

$ python3 main.py pathtodata 

Optionally you can specify a custom batchsize and the number of epochs to train the model. After the model is done training it will be saved as model.pkl.

• Transfer learning on other base models.
• Finetuning base model
• Visualizations