search

yeungchenwa / OCR-SAM

Combining MMOCR with Segment Anything & Stable Diffusion. Automatically detect, recognize and segment text instances, with serval downstream tasks, e.g., Text Removal and Text Inpainting
533 stars 37 forks source link

The effect is not very good #29

Closed Henryplay closed 1 year ago

Henryplay commented 1 year ago

thanks for your contribute! I've used this image 1 and the output is 0 ,there are many characters that were not found.

yeungchenwa commented 1 year ago

Hi! @Henryplay Sorry for late reply. The image you wannna test above is the document image. Our DBNet++ model is mainly trained by secne text images, which is different from the document images. So maybe the model performance on document images is pool in some situations.