Document Layout Analysis on PubLayNet val

15개 결과 · ⬇ CSV · JSON

Overall

Rank	Model	Overall	Text	Title	List	Table	Figure	Paper	Code	Year
1	VGT	0.962	0.950	0.939	0.968	0.981	0.971	Vision Grid Transformer for Document Layout Analysis	alibabaresearch/advancedliteratemachinery	2023
2	TRDLU	0.959	0.958	0.921	0.975	0.976	0.966	Transformer-based Approach for Document Understanding		2022
3	VSR	0.957	0.967	0.931	0.947	0.974	0.964	VSR: A Unified Framework for Document Layout Analysis combining Vision, Semantics and Relations	hikopensource/davar-lab-ocr	2021
3	DETR	0.957	0.947	0.918	0.964	0.981	0.975	Bridging the Performance Gap between DETR and R-CNN for Graphical Object Detection in Document Images		2023
5	LayoutLMv3-B	0.951	0.945	0.906	0.955	0.979	0.970	LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking	huggingface/transformers · microsoft/unilm · pwc-1/Paper-9 · +1	2022
6	DoPTA	0.949	0.944	0.895	0.957	0.977	0.970	DoPTA: Improving Document Layout Analysis using Patch-Text Alignment		2024
6	DiT-L	0.949	0.944	0.893	0.960	0.978	0.972	DiT: Self-supervised Pre-training for Document Image Transformer	huggingface/transformers · microsoft/unilm · thibaultvt/Diard · +1	2022
8	UDoc	0.939	0.939	0.885	0.937	0.973	0.964	Unified Pretraining Framework for Document Understanding		2022
9	ResNext-101-32×8d	0.935	0.930	0.862	0.940	0.976	0.968	Vision Grid Transformer for Document Layout Analysis	alibabaresearch/advancedliteratemachinery	2023
10	DeiT-B	0.932	0.934	0.874	0.921	0.972	0.957	Training data-efficient image transformers & distillation through attention	huggingface/transformers · rwightman/pytorch-image-models · PaddlePaddle/PaddleClas · +37	2020
11	BEiT-B	0.931	0.934	0.866	0.924	0.973	0.957	BEiT: BERT Pre-Training of Image Transformers	huggingface/transformers · rwightman/pytorch-image-models · microsoft/unilm · +11	2021
12	Mask RCNN	0.910	0.916	0.840	0.886	0.960	0.949	PubLayNet: largest dataset ever for document layout analysis	ibm-aur-nlp/PubLayNet · ibm-aur-nlp/PubTabNet · hpanwar08/detectron2 · +3	2019
13	Faster RCNN	0.902	0.910	0.826	0.883	0.954	0.937	PubLayNet: largest dataset ever for document layout analysis	ibm-aur-nlp/PubLayNet · ibm-aur-nlp/PubTabNet · hpanwar08/detectron2 · +3	2019
14	GLAM	0.722	0.878	0.800	0.862	0.868	0.206	A Graphical Approach to Document Layout Analysis	ivanstepanovftw/glam	2023
15	CDeC-Net	–	–	–	–	0.978	–	CDeC-Net: Composite Deformable Cascade Network for Table Detection in Document Images	mdv3101/CDeCNet · samarthramesh/CDeC-Net · 2023-MindSpore-4/Code2	2020