rvl-cdip.

From paper: Visual and Textual Deep Feature Fusion for Document Image Classification

From paper: DocFormer: End-to-End Transformer for Document Understanding

From paper: LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking

From paper: LiLT: A Simple yet Effective Language-Independent Layout Transformer for Structured Document Understanding

From paper: LayoutLMv2: Multi-modal Pre-training for Visually-Rich Document Understanding

From paper: Going Full-TILT Boogie on Document Understanding with Text-Image-Layout Transformer

From paper: DocFormer: End-to-End Transformer for Document Understanding

From paper: LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking

From paper: OCR-free Document Understanding Transformer

From paper: Going Full-TILT Boogie on Document Understanding with Text-Image-Layout Transformer

From paper: LayoutLMv2: Multi-modal Pre-training for Visually-Rich Document Understanding

From paper: LayoutXLM: Multimodal Pre-training for Multilingual Visually-rich Document Understanding

From paper: StrucTexTv2: Masked Visual-Textual Prediction for Document Image Pre-training

From paper: LayoutLM: Pre-training of Text and Layout for Document Image Understanding

From paper: DoPTA: Improving Document Layout Analysis using Patch-Text Alignment

DoPTA: Improving Document Layout Analysis using Patch-Text Alignment. High-resolution (512×512) variant. Table 1. Outperforms Donut-Encoder (93.37%) and StructTexTv2-Small (93.4%) at comparable resolutions.

From paper: DocXClassifier: High Performance Explainable Deep Network for Document Image Classification

HEADoC: Highly Efficient and Accurate Document Classifier Optimized Using Semantic Distances. LARGE variant (90.58M params). Published in Progress in Artificial Intelligence, Oct 2025. Deep attention mechanism fusing textual and visual modalities via semantic distances.

From paper: StrucTexTv2: Masked Visual-Textual Prediction for Document Image Pre-training

From paper: VLCDoC: Vision-Language Contrastive Pre-Training Model for Cross-Modal Document Classification

From paper: GlobalDoc: A Cross-Modal Vision-Language Framework for Real-World Document Image Retrieval and Classification

DoPTA: Improving Document Layout Analysis using Patch-Text Alignment. Standard resolution (224×224) variant. Table 1. Outperforms DiT-L with <1/3rd parameters. 250k pretraining steps on document images.

HEADoC: Highly Efficient and Accurate Document Classifier Optimized Using Semantic Distances. BASE variant (27.7M params). Published in Progress in Artificial Intelligence, Oct 2025. Deep attention mechanism fusing textual and visual modalities via semantic distances.

From paper: Multimodal Side-Tuning for Document Classification

From paper: DiT: Self-supervised Pre-training for Document Image Transformer

From paper: Improving accuracy and speeding up Document Image Classification through parallel systems

From paper: Document Image Classification with Intra-Domain Transfer Learning and Stacked Generalization of Deep Convolutional Neural Networks

From paper: Multimodal Side-Tuning for Document Classification

From paper: DiT: Self-supervised Pre-training for Document Image Transformer

From paper: BEiT: BERT Pre-Training of Image Transformers

From paper: Cutting the Error by Half: Investigation of Very Deep CNN and Advanced Training Strategies for Document Image Classification

From paper: Analysis of Convolutional Neural Networks for Document Image Classification

From paper: Training data-efficient image transformers & distillation through attention

From paper: RoBERTa: A Robustly Optimized BERT Pretraining Approach