Multi-Dataset Experimental Data for Medical Endoscopic Polyp Segmentation

This ScienceDB record provides the reproducibility materials for our study on colorectal polyp segmentation in optical colonoscopy images. Experiments are conducted on five widely used public benchmarks: Kvasir-SEG, CVC-ClinicDB, CVC-ColonDB, ETIS-Larib Polyp DB, and CVC-300 (commonly used as an independent test set and often discussed in relation to the EndoScene / CVC-EndoSceneStill benchmark family).Kvasir-SEG is a curated polyp segmentation benchmark released by Simula and intended for developing and evaluating segmentation and detection methods on colonoscopy images. CVC-ClinicDB is a commonly used set of colonoscopy frames with corresponding ground-truth masks and is distributed via the CVC/Grand-Challenge portal. CVC-ColonDB is released by the CVC Visual Interaction Group as annotated colonoscopy video sequences and frames for polyp analysis. ETIS-Larib Polyp DB provides annotated polyp frames with pixel-level masks and is frequently adopted for cross-dataset generalization evaluation. CVC-300 is widely treated in the literature as a 60-image unseen-domain test set related to EndoScene-based testing protocols.Importantly, the original images and annotations remain subject to the licenses and distribution terms defined by their respective providers. This ScienceDB record does not modify the original licensing conditions; users should obtain the raw datasets from the official sources and cite the corresponding dataset papers or webpages. This record instead shares the experiment protocol and reproducibility package used in our paper, including standardized data organization, train/validation/test split lists, preprocessing settings, and evaluation scripts and/or result files, enabling fair comparison and exact replication across the above benchmarks. Official / Primary Sources (Provenance)· Kvasir-SEG: Simula official dataset page and dataset paper.· CVC-ClinicDB: CVC/UAB databases page and the official distribution portal (e.g., Grand-Challenge).· CVC-ColonDB: CVC Visual Interaction Group official page.· ETIS-Larib Polyp DB: Dataset documentation/registry entry with citation information.· CVC-300: Commonly used 60-image test protocol reported in peer-reviewed literature and associated with EndoScene-based testing.