Text Detection on FUNSD Task 1
0.9543PrecisionDiT-L
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DiT-LBackbone=DiT-L, Detection Framework=Mask R-CNN, Training Strategy=Synthetic dataset (1M images) + fine-tuned, IoU Threshold=0.52022.03 | 0.9543 | 0.9317 | 0.9429 | |
| DiT-BBackbone=DiT-B, Detection Framework=Mask R-CNN, Training Strategy=Synthetic dataset (1M images) + fine-tuned, IoU Threshold=0.52022.03 | 0.9539 | 0.9315 | 0.9425 | |
| DiT-BBackbone=DiT-B, Detection Framework=Mask R-CNN, IoU Threshold=0.52022.03 | 0.947 | 0.9307 | 0.9388 | |
| DiT-LBackbone=DiT-L, Detection Framework=Mask R-CNN, IoU Threshold=0.52022.03 | 0.9452 | 0.9336 | 0.9393 | |
| MAE-BBackbone=MAE-B, Detection Framework=Mask R-CNN, IoU Threshold=0.52022.03 | 0.9441 | 0.9321 | 0.9381 | |
| DeiT-BBackbone=DeiT-B, Detection Framework=Mask R-CNN, IoU Threshold=0.52022.03 | 0.9429 | 0.9237 | 0.9332 | |
| BEIT-BBackbone=BEIT-B, Detection Framework=Mask R-CNN, IoU Threshold=0.52022.03 | 0.9412 | 0.9263 | 0.9337 | |
| ResNeXt-101-32x8dBackbone=ResNeXt-101-32x8d, Detection Framework=Mask R-CNN, IoU Threshold=0.52022.03 | 0.9387 | 0.9229 | 0.9307 | |
| DBNetDetection Framework=DBNet, IoU Threshold=0.52022.03 | 0.8764 | 0.84 | 0.8578 | |
| A Commercial OCR EngineIoU Threshold=0.52022.03 | 0.8762 | 0.826 | 0.8504 | |
| Faster R-CNNDetection Framework=Faster R-CNN, IoU Threshold=0.52022.03 | 0.704 | 0.848 | 0.76 |