AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

M5HisDoc: A Large-scale Multi-style Chinese Historical Document Analysis Benchmark

Yongxin Shi, Chongyu Liu, Dezhi Peng, Cheng Jian, Jiarong Huang, Lianwen Jin

NeurIPS 2023 · Image generation › Autoregressive and latent generation · approach: Retrieval and matching with generative features · 6 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
raise IoU threshold from 0.5 to 0.7DBNet++ · M5HisDoc-H text line detectionF1-score: 90.77 -> 38.57Hurt
raise IoU threshold from 0.5 to 0.7PAN · M5HisDoc-H text line detectionF1-score: 86.22 -> 71.55Hurt
raise IoU threshold from 0.5 to 0.7TextSnake · M5HisDoc-H text line detectionF1-score: 90.85 -> 84.76Mixed
raise IoU threshold from 0.5 to 0.7YOLOX · M5HisDoc-H character detectionF1-score: 97.35 -> 84.76Hurt
pass detected text lines through CRNN for end-to-end evaluationMask R-CNN · M5HisDoc-HF1-score vs 1-NED: 84.36 -> 66.17Hurt
Cascade R-CNN instead of Mask R-CNNregression-based detector · M5HisDoc-H text line detection IoU 0.5F1-score: 84.36 -> 86.74Improved
CTC-based CRNN instead of attention-based ASTERtext line recognizer · M5HisDoc-Hcorrect rate: 86.95 -> 90.22Improved
train with character bounding box annotations (Peng et al.*) (ablation)Peng et al. · M5HisDoc-H text line recognitioncorrect rate: 88.16 -> 89.97Improved
heuristic reading order instead of Augmented XY Cutreading order prediction · M5HisDoc-HARD: 14.63 -> 5.27Improved
evaluate on distorted M5HisDoc-H instead of M5HisDoc-RLayoutReader · reading order predictionARD: 5.49 -> 8.68Hurt
ViT backbone instead of SwinTransformercharacter recognizer · M5HisDoc-Htop-1 accuracy: 94.68 -> 94.09Hurt
measure macro accuracy instead of top-1 accuracyResNet50 · M5HisDoc-H character recognitionaccuracy: 94.40 -> 69.61Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy