Fabric Defect Detection With Deep Learning and False Negative Reduction

Quality control is an area of utmost importance for fabric production companies. By not detecting the defects present in fabrics, companies are at risk losing money and reputation with a damaged product. In traditional system, inspection accuracy 60-75% observed. order to reduce these costs, fast automatic defect detection which can be complemented operator decision, proposed this paper. To perform task detection, custom Convolutional Neural Network (CNN) was used work. obtain well-generalized training process, more than 50 types were used. Additionally, as undetected (False Negative - FN) usually has higher cost company non-defective being classified defective one (false positive), FN reduction methods system. testing, when system mode, average 75% attained; however, if method applied, intervention operator, 95% achieved. These results demonstrate ability detect many different good whilst faster computationally simple.