Document Layout Analysis on PubLayNet (val)
0.978mAPCDeC-Net++
Evaluation Results
| Method | Links | ||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| CDeC-Net++Training Dataset=IIIT-AR-13K, Training #Image=9K, Fine-tuning Dataset=PubLayNet, Fine-tuning #Image=340K, IoU=0.5-0.92020.08 | 0.978 | 0.975 | 0.993 | 0.984 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CDeC-NetTraining Dataset=PubLayNet, Training #Image=340K, IoU=0.5-0.92020.08 | 0.967 | 0.97 | 0.988 | 0.978 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| M-RCNNTraining Dataset=PubLayNet, Training #Image=340K, IoU=0.5-0.92020.08 | 0.96 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VSRdataset_split=val2021.05 | 0.957 | — | — | — | 0.967 | 0.931 | 0.947 | 0.974 | 0.964 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| StrucTexTv2Model size=Large2023.03 | 0.955 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| F-RCNNTraining Dataset=PubLayNet, Training #Image=340K, IoU=0.5-0.92020.08 | 0.954 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| StrucTexTv2Model size=Small2023.03 | 0.954 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LayoutLMv3_BASEFramework=Cascade R-CNN, Backbone=Transformer2022.04 | 0.951 | — | — | — | — | — | — | — | — | 0.945 | 0.906 | 0.955 | 0.979 | 0.97 | — | — | — | — | — | — | — | — | — | — | — | — | |
| Layout LMv3Backbone size=Base2023.03 | 0.951 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DiTBackbone size=Large2023.03 | 0.949 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DiT_BASEFramework=Cascade R-CNN, Backbone=Transformer2022.04 | 0.945 | — | — | — | — | — | — | — | — | 0.944 | 0.889 | 0.948 | 0.976 | 0.969 | — | — | — | — | — | — | — | — | — | — | — | — | |
| PP-StructureV2Backbone=PP-PicoDet, Speed (ms)=41.22022.10 | 0.94 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| UDocFramework=Faster R-CNN, Backbone=ResNet-502022.04 | 0.939 | — | — | — | — | — | — | — | — | 0.939 | 0.885 | 0.937 | 0.973 | 0.964 | — | — | — | — | — | — | — | — | — | — | — | — | |
| UDoc2023.03 | 0.939 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DiT_BASEFramework=Mask R-CNN, Backbone=Transformer2022.04 | 0.935 | — | — | — | — | — | — | — | — | 0.934 | 0.871 | 0.929 | 0.973 | 0.967 | — | — | — | — | — | — | — | — | — | — | — | — | |
| BEITBackbone size=Base2023.03 | 0.931 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Mask RCNNdataset_split=val2021.05 | 0.91 | — | — | — | 0.916 | 0.84 | 0.886 | 0.96 | 0.949 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PubLayNetFramework=Mask R-CNN, Backbone=ResNet-1012022.04 | 0.91 | — | — | — | — | — | — | — | — | 0.916 | 0.84 | 0.886 | 0.96 | 0.949 | — | — | — | — | — | — | — | — | — | — | — | — | |
| Faster RCNNdataset_split=val2021.05 | 0.902 | — | — | — | 0.91 | 0.826 | 0.883 | 0.954 | 0.937 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| layoutparserBackbone=Detectron2, Speed (ms)=2900.02022.10 | 0.8898 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CDeC-Net++Training Dataset=IIIT-AR-13K, Training #Image=9K, IoU=0.5-0.92020.08 | 0.734 | 0.767 | 0.785 | 0.776 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| [46]Architecture=Mask R-CNN2022.03 | — | — | — | 0.91 | 0.916 | 0.84 | 0.886 | 0.96 | 0.949 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BEiT-BBackbone=BEiT-B, Architecture=Mask R-CNN2022.03 | — | — | — | 0.931 | 0.934 | 0.866 | 0.924 | 0.973 | 0.957 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BEiT-B2024.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 0.934 | 0.866 | 0.924 | 0.973 | 0.957 | 0.931 | |
| DeiT-BBackbone=DeiT-B, Architecture=Mask R-CNN2022.03 | — | — | — | 0.932 | 0.934 | 0.874 | 0.921 | 0.972 | 0.957 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeiT-B2024.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 0.934 | 0.874 | 0.921 | 0.972 | 0.957 | 0.932 | |
| DiT-BBackbone=DiT-B, Architecture=Mask R-CNN2022.03 | — | — | — | 0.935 | 0.934 | 0.871 | 0.929 | 0.973 | 0.967 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DiT-BParameters=87M2024.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 0.944 | 0.889 | 0.948 | 0.976 | 0.969 | 0.945 | |
| DiT-B (C)Backbone=DiT-B, Architecture=Cascade Mask R-CNN2022.03 | — | — | — | 0.945 | 0.944 | 0.889 | 0.948 | 0.976 | 0.969 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DiT-LBackbone=DiT-L, Architecture=Mask R-CNN2022.03 | — | — | — | 0.941 | 0.937 | 0.879 | 0.945 | 0.974 | 0.968 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DiT-LParameters=304M2024.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 0.944 | 0.893 | 0.96 | 0.978 | 0.972 | 0.949 | |
| DiT-L (C)Backbone=DiT-L, Architecture=Cascade Mask R-CNN2022.03 | — | — | — | 0.949 | 0.944 | 0.893 | 0.96 | 0.978 | 0.972 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Donut-EncoderParameters=72M2024.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 0.939 | 0.875 | 0.952 | 0.976 | 0.969 | 0.942 | |
| DOPTAParameters=85M2024.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 0.944 | 0.895 | 0.957 | 0.977 | 0.97 | 0.949 | |
| GLAM2023.08 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 0.878 | 0.8 | 0.862 | 0.868 | 0.206 | 0.722 | — | — | — | — | — | — | |
| LayoutLM v32023.08 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 0.945 | 0.906 | 0.955 | 0.979 | 0.97 | 0.951 | — | — | — | — | — | — | |
| LayoutLMv3-BaseParameters=133M, OCR as input=true, Longer finetuning schedule=true2024.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 0.945 | 0.906 | 0.955 | 0.979 | 0.97 | 0.951 | |
| M2DocOCR as input=true2024.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 0.943 | 0.887 | 0.952 | 0.973 | 0.967 | 0.945 | |
| MAE-BBackbone=MAE-B, Architecture=Mask R-CNN2022.03 | — | — | — | 0.93 | 0.933 | 0.865 | 0.918 | 0.973 | 0.959 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MAE-B2024.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 0.933 | 0.865 | 0.918 | 0.973 | 0.959 | 0.93 | |
| MBC2023.08 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 0.888 | 0.528 | 0.881 | 0.977 | 0.84 | 0.823 | — | — | — | — | — | — | |
| ResNextBackbone=ResNext-101-32x8d, Architecture=Mask R-CNN2022.03 | — | — | — | 0.92 | 0.916 | 0.845 | 0.918 | 0.971 | 0.952 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ResNeXt2024.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 0.916 | 0.845 | 0.918 | 0.971 | 0.952 | 0.92 | |
| ResNext (C)Backbone=ResNext-101-32x8d, Architecture=Cascade Mask R-CNN2022.03 | — | — | — | 0.935 | 0.93 | 0.862 | 0.94 | 0.976 | 0.968 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| StructTextv2-LargeParameters=238M, OCR as input=true, Longer finetuning schedule=true2024.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 0.955 | |
| UDoc2024.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 0.939 | 0.885 | 0.937 | 0.973 | 0.964 | 0.939 | |
| VGTParameters=174M, OCR as input=true2024.12 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 0.948 | 0.928 | 0.953 | 0.977 | 0.967 | 0.955 | |
| VSR2023.08 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 0.967 | 0.931 | 0.947 | 0.974 | 0.964 | 0.957 | — | — | — | — | — | — | |
| YOLO v5x62023.08 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 0.917 | 0.791 | 0.869 | 0.965 | 0.947 | 0.898 | — | — | — | — | — | — |