CLIP From Scratch

This is my implementation of the CLIP (Contrastive Language-Image Pretraining) model.

---

Project Structure

• clip_deployment/
  Contains a Gradio app that lets you upload an image and returns the top 5 matching captions from a pre-stored captions list.

• CLIP_training.ipynb
  A Jupyter notebook where the entire training process of the CLIP model is implemented.

---

How to Use

1. Training
   Open and run the CLIP_training.ipynb notebook to train the CLIP model on your dataset.

2. Deployment
   Inside the clip_deployment folder, launch the Gradio app to interact with the trained model by uploading images and getting the most relevant captions.

---

Feel free to explore, experiment on this project!