Comic Book Strip OCR and Text Generation

Authors' explanation here.

1. Clone repo.
2. Create a python3 venv using python3 -m venv env and then source env/bin/activate or <venv>\Scripts\activate.bat, depending on your OS.
3. Run make install to download the dependencies.
4. Run export COHERE_APIKEY=REPLACE, replacing REPLACE with your Cohere API key.
5. Run app.py (with --flagging if you want to allow users to flag generations as incorrect, offensive, etc.), and you now have a local Gradio app!

• Project for the TAMU Datathon to solve the CBRE Challenge: Get in Line and the Best Use of NLP by Cohere - MLH challenges.
• When evaluated with difflib.SequenceMatcher on the challenge-provided files found here, the OCR model achieves ~89% average accuracy for all the example comic strips.
• Since we are open-sourcing the necessary code for others to create their own Gradio link, please be mindful of what data you submit to others' links and the privacy of the data people submit to your links.
• Check out the Devpost for more info.

• Tesseract for their OCR engine.
• Cohere.ai for their text generation API.