HFID-AL: A Hyperspectral Facial Imaging Dataset for Acne Lesion Detection and Segmentation

HFID-AL is a curated hyperspectral facial imaging dataset for acne lesion detection and segmentation. The dataset contains approximately 2,900 hyperspectral image cubes collected from 300 participants using an LCTF-based imaging system covering wavelengths from 400–1000 nm with 121 spectral bands at 5 nm intervals. All images were standardized to 512 × 512 resolution. Among them, 744 images were annotated with expert pixel-level acne lesion masks by board-certified dermatologists in COCO format. The dataset supports research in:- hyperspectral dermatological imaging- acne lesion segmentation- spectral feature analysis- medical image understanding- AI-based skin analysis Released resources include:- hyperspectral image cubes- annotation masks- metadata- preprocessing scripts- benchmark code- documentation Baseline experiments using U-Net, LWB-UNet, Mamba-UNet, and VML-UNet are also included.