ECCV2022/PARSeq-OCR

Use PyTorch inference mode

Fix and update dependencies

Add more demo images

Remove toggle for detailed output

UI improvements

Add clarification about handwritten text datasets

Fix __call__ return signature when image is None

Show raw output with confidence values + fix layout

Handle empty image input

Better formatting

Clearer instructions

Fix paths of demo images

Add app, config, and data

initial commit