how can I test this model ??
#1
by
RAyyyyyEN
- opened
I see that this model has good CER and WER, but I want to test whether it can really extract Arabic handwritten text in a real-world context
RAyyyyyEN
changed discussion title from
how can test this model ??
to how can I test this model ??
I trained it only on KHATT_v1.0 dataset. You can upload an Arabic handwritten image and test it here: https://huggingface.co/spaces/johnlockejrr/PyLaia-ar-hand_v1
You have to provide line level images (https://huggingface.co/datasets/johnlockejrr/KHATT_v1.0_dataset). So, if you want to use it on a page level, you can make a pipeline using YOLO to segment your image, detect the lines, cut them and send to PyLaia for recognition.