DentFilm v1.0: an expert-annotated clinical image dataset of complete denture biofilm

Publicly available expert-annotated datasets are essential for developing robust artificial intelligence (AI) models in dentistry. Although several image datasets have been developed for applications involving natural teeth, no publicly available dataset currently exists for automated analysis of biofilm on conventional acrylic resin complete dentures. To address this gap, we present the first expert-annotated clinical image dataset specifically designed for computer vision research on complete denture biofilm. The dataset comprises 720 total files (630 standardized clinical photographs and 90 expert-annotated segmentation masks) obtained from 30 edentulous participants wearing maxillary conventional acrylic resin complete dentures. Images were acquired under controlled conditions at three stages of a denture hygiene protocol (before biofilm disclosure, after 0.5% basic fuchsin application, and after biofilm removal) and include frontal, right lateral, left lateral, right oblique, left oblique, occlusal, and intaglio views. Expert-generated pixel-level segmentation masks were manually created using the disclosed images as visual references, and each image is accompanied by structured metadata describing participant identification, acquisition stage, photographic view, and annotation files. The dataset is organized using a standardized directory structure to facilitate reproducible machine learning workflows and is publicly available through Zenodo. This resourceprovides the first publicly available benchmark dataset for complete denture biofilm analysis, supporting semantic segmentation, image classification, explainable AI, algorithm benchmarking, and the development of automated systems for denture hygiene assessment.