Document-level OCR on FUNSD 50 images (test)
67.14F1 ScoreUNIT
Evaluation Results
| Method | Links | |
|---|---|---|
| UNITBackbone=ViT-H/14, #Params=632M, Input=896 x 8962024.09 | 67.14 | |
| NougatBackbone=SAM-ViT-B, #Params=247M, Input=896 × 6722024.09 | 55.35 | |
| RADIOBackbone=ViT-H/14, #Params=632M, Input=896 x 896, Reimplemented=true2024.09 | 26.12 | |
| VaryBackbone=SAM-ViT-B, #Params=525M, Input=1024 x 1024, Reimplemented=true2024.09 | 21.01 | |
| DonutBackbone=Swin-B, #Params=260M, Input=1280 × 9602024.09 | 9.08 |