Fab-IL

This dataset is designed for research in continual learning (also known as incremental learning) within the industrial visual inspection domain. It comprises fabric surface images collected from four distinct textile manufacturing factories, each representing a unique data domain due to variations in equipment, raw materials, and fabric types. The core challenge is to enable a model to learn from these domains sequentially without catastrophically forgetting previously acquired knowledge. Total Samples: 97,575 RGB images Number of Domains (Factories): 4 Domains: Factory_tx, Factory_sy, Factory_zh, Factory_yc . Task: Multi-class classification for defect detection. Number of Classes: 16 defect types + 1 "normal" class (defect-free).