4 ms·
A different approach from vanilla OCR/parsing seems to be this mixed ColPali integrating a purposed small vision models and a ColBERT type indexing for retrieva
by jimmySixDOF 2y ago
A different approach from vanilla OCR/parsing seems to be this mixed ColPali integrating a purposed small vision models and a ColBERT type indexing for retrieval. So - if search is the intended use case - it can skip the whole OCR step entirely.
[1] https://huggingface.co/blog/manu/colpali https://huggingface.co/blog/manu/colpali