steel defect

NEU-DET is a steel plate surface defect dataset launched by Northeastern University, which includes 1800 sheets of 200 × 200 pixels and 6 common defects; GC10-DET is an extended dataset for real industrial scenarios, covering 10 types of metal surface defects, with larger image scales and more complex working conditions, which is suitable for practical implementation needs; Constructing synthetic defect data based on improved DCGAN generation technology, which can flexibly expand samples, balance defect categories, and enhance model generalization ability.