A natural text to image search using CLIP
This is an example that implements search for War Images (currently for Ukraine War only)
Features
- You'll be able to use natural text queries for image search
- The search will work even when the image has no such info in the captions or metadata
- The search will work even when you make a
typpotypo
Example natural text queries and results
Possibilities are endless. As long as the data has the images matching your natural text, you get sruprisingly accurate results.
Checkout the code
Just run the code in Kaggle notebook or copy the notebook to your favorite ML notebook editor e.g. Colab, Jupyter, etc.
- Backend to search images with natural text (powered by CLIP and Jina CLIP-as-service)
- Demo on Kaggle
- REST APIs for search to be used in frontend
- Frontend Web UI to search easily
- Backend APIs to upload image to the dataset
- Frontend Web UI to upload your images to the dataset(with source citation)
- Frontend Web UI to change the image dataset
- Production deployment guide
- Show image match score
- Command palette for quick actions
You can contribute to this project in two ways
- Contributing to the code to refine this example and add features
- Contributing to the dataset for war images. No labeling required.
MIT