Image classification based on the "Fashion MNIST" dataset for the Machine Learning course
- Documentation website https://leonidaszotos.github.io/FashionMNIST_G25/index.html
- Download the data from https://www.kaggle.com/zalando-research/fashionmnist
- Put ALL of the files in the data folder. All the .csv and ubyte stuff. Or nothing will happen!
- in load_data
- subset : Set to any integer or None for full data
- in multi_model_run
- reduce_dims : name of reduction method
- model_list : pass in a list of any models you want with their params
- metrics : list of all metrics
- folds : no of folds for k folds
- All the configurations and main running stuff : runner.py
- I patched in a way to just take a subset of the data for faster computation : Just go to runner.py, in the load_data function there just add whatever number of images you want for subset. Eg 1000.
- All the functions : backbone.py
- The .ipynb jupyter notebooks are auto generated but you can use them to test new features etc
- They WILL be overwritten. So make sure to either rename it to something else or save them elsewhere to be safe.
- First install all the requirements using
pip install -r requirements.txt - python3 runner.py should run everything youve set up
- If you are using WSL/Linux : please first do
- chmod +x pusher.sh (Just once)
- Everytime you want to save to github :
./pusher.sh "commit_message"- This runs a bunch of code formatting stuff in the background