GastroEndoNet

GastroEndoNet is a curated gastrointestinal endoscopy image dataset containing 7,499 JPG images organized into six clinically meaningful classes: Cancer (402), Gerd (996), Gerd Normal (1,132), Polyp (959), Polyp Normal (3,565), and Spot (445). The dataset is released in its original class-wise format to support immediate use in medical computer vision research, including classification, benchmarking, and dataset engineering. It provides a practical mix of disease and normal-control categories, enabling both lesion-focused modeling and normal-vs-abnormal learning in a single unified resource. Image sizes are broadly consistent (around 550 x 512 pixels on average), making the collection convenient for standardized preprocessing and reproducible pipelines.

Folder structure:
GastroEndoNet/
Cancer/ (402 JPG images)
Gerd/ (996 JPG images)
Gerd Normal/ (1,132 JPG images)
Polyp/ (959 JPG images)
Polyp Normal/ (3,565 JPG images)
Spot/ (445 JPG images)

This release focuses on the six raw class folders as the core dataset asset, giving researchers full flexibility to apply their own split strategies, augmentation policies, and evaluation protocols. It serves as a solid baseline dataset for GI image analysis and model development.