Table 8 Summary of public colorectal neoplasm endoscopy datasets
From: Improving generalization of polyp detection via conditional StyleGAN augmented training
Dataset | Source | Content | Qty. |
|---|---|---|---|
Kvasir-SEG79 | SimulaMet (Norway) | Polyp images with pixel-wise segmentation masks. | 1000 |
CVC-ClinicDB80 | CVC (Spain) | Frames from colonoscopy videos with segmentation masks. | 612 |
CVC-ColonDB81 | CVC (Spain) | Frames from short videos with segmentation masks. | 380 |
ETIS-Larib82 | ETIS & Larib (FR/TN) | Images of varied quality, including difficult cases. | 196 |
HyperKvasir83 | SimulaMet (Norway) | Massive, multi-class dataset of upper/lower GI findings. | >110k |
PICCOLO84 | Spanish Hospitals | High definition NBI images with histopathological labels. | >3.4k |
SUN DB81 | SUNY Buffalo (USA) | Full-length colonoscopy videos with detailed polyp annotations. | 99 |
ASU-Mayo DB85 | ASU & Mayo Clinic | Colonoscopy videos from the 2015 MICCAI Challenge. | >38k |
