Spaces:
Running
Add paddle-calfa-tiny
Calfa's open Armenian recognition model paddle-calfa-tiny (based on PP-OCRv6-tiny), from the open OCR package (https://github.com/calfa-co/hye-open-ocr), with the free Calfa Vision line segmenter.
Evaluation output (dataset evaluation_kit, 197 pages, no post-processing): https://huggingface.co/datasets/CalfaUser/paddle-calfa-tiny-acop-eval/tree/main/ACoPPer_paddle-calfa-tiny
Thank you for your interest in our published benchmark and for submitting a pull request to add your model to the leaderboard!
Could you please share the configuration details and settings you used for your model? This would help us reproduce your results and ensure a fair comparison with other models on the leaderboard.
We will add your hy-open-ocr model to the ACoPPer leaderboard. As part of this update, we also plan to rerun and update the results for some existing models to reflect recent updates.
Thank you again for your contribution!
We have used the hye-open-ocr package (https://github.com/calfa-co/hye-open-ocr) and more specifically the open paddle model it provides: paddle-calfa-tiny (https://github.com/calfa-co/hye-paddle), with pages limited to 3000px in height, and max detections put to 5,000.