Document Layout Analysis on DocLayNet
79.7mAPDocLayout-YOLO
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DocLayout-YOLOBackbone=v10m++, Pretrain Dataset=DocSynth, 300K2024.10 | 79.7 | 93.4 | |
| DINO-4scaleBackbone=R50, Pretrain Dataset=ImageNet1K2024.10 | 77.7 | 93.5 | |
| YOLO-v10Backbone=v10m2024.10 | 76.2 | 93 | |
| LayoutLMv3-BBackbone=ViT-B, Pretrain Dataset=IIT-CDIP, 42M2024.10 | 75.4 | 92.1 | |
| DAVELLM Decoder=Llama-3.2-3B-Instruct2025.12 | 74.1 | — | |
| DiT-Cascade-BBackbone=ViT-B, Pretrain Dataset=IIT-CDIP, 42M2024.10 | 73.2 | 87.6 | |
| DiT-Cascade-LBackbone=ViT-L, Pretrain Dataset=IIT-CDIP, 42M2024.10 | 72.6 | 84.9 | |
| SigLIP 2LLM Decoder=Llama-3.2-3B-Instruct2025.12 | 70.8 | — | |
| AIMv2LLM Decoder=Llama-3.2-3B-Instruct2025.12 | 70.5 | — | |
| DinoV2LLM Decoder=Llama-3.2-3B-Instruct2025.12 | 68.4 | — | |
| MAE-WebLLM Decoder=Llama-3.2-3B-Instruct2025.12 | 64.6 | — | |
| Pix2StructLLM Decoder=Llama-3.2-3B-Instruct2025.12 | 56.7 | — | |
| DolphinLLM Decoder=Llama-3.2-3B-Instruct2025.12 | 53.8 | — |