sidmo

3 ms·

1. 1.▲
   by sidmo 2y ago

   I'd recommend checking out vision language models. They generate embeddings of the images themselves (as a collection of patches) and you can see query matching displayed as a heatmap over the document. Picks up text that OCR misses. I
2. 2.▲
   by sidmo 2y ago

   VLMs are cool - they generate embeddings of the images themselves (as a collection of patches) and you can see query matching displayed as a heatmap over the document. Picks up text that OCR misses. Here's an open-source API demo I bui
3. 3.▲
   by sidmo 2y ago

   If you are looking for the latest/greatest in file processing i'd recommend checking out vision language models. They generate embeddings of the images themselves (as a collection of patches) and you can see query matching display