ORGANISATION

RVL-CDIP

400,000 scanned document images labeled across 16 categories, used for document image classification tasks.

About Dataset

RVL-CDIP is a large dataset of 400,000 scanned document images, each labeled into one of 16 categories such as letters, forms, invoices, and scientific reports. Derived from the IIT-CDIP document collection, the dataset provides a standardized benchmark for document image classification, capturing the visual and structural variety found in real-world scanned business and legal documents.

Purpose of Dataset

Rvl-cdip Is Used To Train And Evaluate Models For Document Image Classification, Layout Analysis, And Document Type Recognition. It Supports Research In Document Ai Systems Used For Automated Document Processing, Archival Organization, And Enterprise Content Management. The Dataset Is A Common Benchmark For Comparing Convolutional And Transformer-based Document Classification Architectures.