OCR recognition doesn't seem to have mask generation, which I think would be a welcome feature

kijai / ComfyUI-Florence2

Inference Microsoft Florence2 VLM

MIT License

768 stars 51 forks source link

Closed supertaog closed 2 months ago

supertaog commented 4 months ago

Maybe it's my problem, but I've tried, and the mask is always a black picture with nothing.

Oceanswave commented 2 months ago

same. outputting a mask of all text with ocr_with_regions would be really useful.