A Controlled Smartphone Image Dataset for AI-Based Workplace Illuminance Estimation

Version 3 notes: Corrected surface_group for 494 miscategorized rows; fixed label_kind for 935 rows (1point→5point); fixed malformed surface column; reclassified 467 white_paper images from colored_paper group; synced surface_group_norm; normalized 24 table_lc### surface names to 'table'. All lux measurements and image features unchanged. Dataset of 5,814 labeled smartphone photographs of indoor work surfaces collected under controlled illumination for training and evaluation of machine-learning models for workplace illuminance (lux) estimation. Images were captured with a Samsung Galaxy A16 at fixed camera settings (ISO 50, 1/50 s exposure, 4000 K white balance, 45° angle, 31 cm height). Ground-truth lux values were measured with a Lutron LX-101A digital lux meter at one central point or at five spatial positions per image. Surface types include bare table surfaces, white paper, 22 colors of paper, and other surfaces. Illuminance ranges from 2 to 2,280 lux. The record includes image ZIP files, the master annotation table (annotation_master_latest.csv; 5,814 rows × 23 columns), benchmark CSV files, and the full-dataset feature table (feature_master_latest.csv; 5,814 rows × 133 columns).