HyperAI초신경

홈 뉴스 연구 논문 튜토리얼 데이터셋 백과사전 SOTA LLM 모델 GPU 랭킹 컨퍼런스

한국어

HyperAI초신경

Document Layout Analysis On Publaynet Val

평가 지표

Figure

List

Overall

Table

Text

Title

평가 결과

이 벤치마크에서 각 모델의 성능 결과

모델 이름	Figure	List	Overall	Table	Text	Title	Paper Title	Repository
DETR	0.975	0.964	0.957	0.981	0.947	0.918	Bridging the Performance Gap between DETR and R-CNN for Graphical Object Detection in Document Images	-
VSR	0.964	0.947	0.957	0.974	0.967	0.931	VSR: A Unified Framework for Document Layout Analysis combining Vision, Semantics and Relations
DoPTA-HR	0.970	0.957	0.949	0.977	0.944	0.895	DoPTA: Improving Document Layout Analysis using Patch-Text Alignment	-
ResNext-101-32×8d	0.968	0.940	0.935	0.976	0.930	0.862	Vision Grid Transformer for Document Layout Analysis
UDoc	0.964	0.937	0.939	0.973	0.939	0.885	Unified Pretraining Framework for Document Understanding	-
LayoutLMv3-B	0.970	0.955	0.951	0.979	0.945	0.906	LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking
GLAM	0.206	0.862	0.722	0.868	0.878	0.800	A Graphical Approach to Document Layout Analysis
CDeC-Net	-	-	-	0.978	-	-	CDeC-Net: Composite Deformable Cascade Network for Table Detection in Document Images
DiT-L	0.972	0.960	0.949	0.978	0.944	0.893	DiT: Self-supervised Pre-training for Document Image Transformer
TRDLU	0.966	0.975	0.959	0.976	0.958	0.921	Transformer-based Approach for Document Understanding	-
Faster RCNN	0.937	0.883	0.902	0.954	0.910	0.826	PubLayNet: largest dataset ever for document layout analysis
Mask RCNN	0.949	0.886	0.910	0.960	0.916	0.840	PubLayNet: largest dataset ever for document layout analysis
DeiT-B	0.957	0.921	0.932	0.972	0.934	0.874	Training data-efficient image transformers & distillation through attention
VGT	0.971	0.968	0.962	0.981	0.950	0.939	Vision Grid Transformer for Document Layout Analysis
BEiT-B	0.957	0.924	0.931	0.973	0.934	0.866	BEiT: BERT Pre-Training of Image Transformers

0 of 15 row(s) selected.