Grayscale Document Images Across 16 Categories

The RVL-CDIP dataset contains 400,000 grayscale images of documents, evenly distributed across 16 classes with 25,000 images per class. It is partitioned into 320,000 training images, 40,000 validation images, and 40,000 test images. Images are sized so their largest dimension does not exceed 1000 pixels.

Use Cases

• Train a multi-class image classifier to predict the 16 document categories, such as 'advertisement' or 'budget'.
• Benchmark document image classification models using the provided 320,000/40,000/40,000 train/validation/test split.
• Pre-train a vision model on 400,000 grayscale document images for transfer learning to related tasks.
• Analyze model performance across the 16 balanced classes, each containing 25,000 samples.

Strengths

• Large scale of 400,000 total grayscale document images.
• Balanced class distribution with exactly 25,000 images in each of the 16 categories.
• Standardized image preprocessing with a maximum dimension of 1000 pixels.
• Explicit data splits provided: 320,000 for training, 40,000 for validation, and 40,000 for testing.

Limitations

• Images are grayscale, lacking color information that may be present in original documents.
• The dataset is focused on document images, which may limit generalizability to other visual domains.
• Specific geographic or temporal coverage of the source documents is not detailed.

Provenance

Source

Ryerson Vision Lab (RVL), via a Hugging Face repository maintained by 'chainyo'.

Collection Method

null

Time Range

null

Freshness

null

Geography

null