Research map / Vision-language and multimodal / Human and embodied modeling
Vision Grid Transformer for Document Layout Analysis
Cheng Da, Chuwei Luo, Qi Zheng, Cong Yao
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add pre-trained Grid Transformer stream (text grid) to ViT (ablation) | VGT · DocBank2K | mAP: 59.61 -> 65.94 | Improved |
| pre-train GiT with MGLM and SLM (ablation) | VGT · DocBank2K | mAP: 64.01 -> 65.94 | Improved |
| pre-train GiT with MGLM and SLM (ablation) | VGT · PubLayNet2K | mAP: 87.76 -> 88.44 | Improved |
| add text tokens to GiT-only model (ablation) | GiT only · DocBank2K | mAP: 40.56 -> 49.15 | Improved |
| double image input with a second ViT instead of GiT (ablation) | VGT · DocBank2K | mAP: 59.61 -> 58.57 | No clear effect |
| add pre-trained GiT to ResNeXt-101 backbone | Cascade R-CNN ResNeXt-101 · DocBank2K | mAP: 57.04 -> 63.05 | Improved |
| use LayoutLM instead of BERT embeddings in grid (ablation) | VGT · PubLayNet2K / DocBank2K | mAP: 87.97 -> 87.76 / 63.97 -> 64.01 | No clear effect |
| add SLM to MGLM pre-training (ablation) | GiT pre-training · DocBank2K | mAP: 64.539 -> 65.167 | Improved |
| use SLM only instead of MGLM only (ablation) | GiT pre-training · DocBank2K | mAP: 64.539 -> 65.112 | Improved |
| two-stream design increases parameters and latency | VGT · inference | ms: 210 ms -> 460ms | Hurt |
Closest papers by meaning
- DocFormerv2: Local Features for Document Understanding (AAAI 2024 · similarity 0.87)
- A Simple yet Effective Layout Token in Large Language Models for Document Understanding (CVPR 2025 · similarity 0.86)
- Graph-based Document Structure Analysis (ICLR 2025 · similarity 0.85)
- RoDLA: Benchmarking the Robustness of Document Layout Analysis Models (CVPR 2024 · similarity 0.85)
- LayoutLLM: Layout Instruction Tuning with Large Language Models for Document Understanding (CVPR 2024 · similarity 0.84)
- M2Doc: A Multi-Modal Fusion Approach for Document Layout Analysis (AAAI 2024 · similarity 0.84)
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.
