Dataset Description Nicotine_stainV1 is a novel benchmark dataset developed to support research on automated nicotine-stain segmentation from human fingertips. Due to the lack of publicly available datasets in this domain, Nicotine_stainV1 was constructed entirely from scratch. The initial collection phase gathered over 300 raw fingertip images from diverse online sources, including Google Image Search and various social media platforms. To further increase the heterogeneity and robustness of the dataset, synthetic samples were generated using state-of-the-art vision-language generative models (Sora-2 and Nano-Banana). This acquisition strategy ensures broad coverage of variations in hand pose, illumination, background complexity, and skin tone. All images were manually annotated at the pixel level using the Roboflow platform to produce high-quality segmentation masks that precisely delineate nicotine-stained regions. To mitigate overfitting and enhance generalization, extensive data augmentation was applied, including rotations (90° and random angles between −15° and +15°), shearing (±10%), and photometric adjustments (±15% brightness). The final dataset contains 743 image-mask pairs. The dataset is organized into training, validation, and testing splits following an 80/10/10 stratified sampling strategy to preserve class balance across all partitions. Folder structure: dataset/ train/ images/ masks/ valid/ images/ masks/ Nicotine_stainV1 represents the first publicly released dataset dedicated to nicotine-stain segmentation and aims to facilitate reproducible research in clinical imaging, digital dermatology, and computational health monitoring.