awhb/art-style-classifier

Art Style Classifier

Classify your art by a fine art movement/style! Especially for those new to fine art history (like myself). Do read up more on the art styles/movements by visiting WikiArt.

A HuggingFace Space which was built using the FastAI library following the FastAI course. Training dataset of around 4000 images obtained from WikiArt by user Sivar Azadi, reposted as a Kaggle Dataset.

Do checkout my process of fine-tuning the ConvNeXt-small-22k model using the fastai library on this Kaggle Notebook. Leave an upvote there if you enjoy!

Tools/Technologies Used

• —fastai library (and dependencies)
• —Gradio (version 6.14.0)
• —huggingface-cli for authentication/deployment
• —For Python libraries view requirements.txt

Project File Structure

├── LICENSE
├── README.md
├── app.ipynb
├── app.py
├── images
│   ├── baroque_art.jpg
│   ├── realism_art.jpg
│   ├── renaissance_art.jpg
│   └── romanticism_art.jpg
├── model.pkl
└── requirements.txt

Instructions to deploy to HF spaces

(Note: Current local repo should already be tracking a HF remote space)

Part 1: Init Git LFS

1. 1.Run git lfs install after installing git-lfs (OS-dependent, I used Homebrew)
2. 2.Run git lfs track "*.jpg".
3. 3.Run git add .gitattributes, git commit -m "Track images with Git LFS".
4. 4.Run git add ... for all image files.
5. 5.Run git commit -m "Add art images tracked by Git LFS"

Part 2: Deploy to HF-Spaces

1. 1.Run git push.

Additional Note:

My local repository sets remotes of origin for the HuggingFace space and upstream for the GitHub repository. This behaviour can be modified, depending on where you begin initialising from.