CTPN-based cloth defect detection method

The invention discloses a CTPN-based cloth defect detection method. The method comprises the following steps: S1, acquiring image information of a cloth data set; s2, dividing image samples of the cloth data set; s3, feature extraction of images of the cloth data set; s4, CTPN-based model construction of a cloth data set: (1) an image preprocessing module; (2) optimizing a network framework: replacing the network with a Mobilenetv2 network to serve as a feature extractor; (3) using a bidirectional LSTM (Long Short Term Memory); (4) a specific anchor design module; (5) a post-processing module; and S5, defect detection of the cloth data set. The method is stable, reliable and high in generalization ability, can be directly popularized, completely abandons a traditional algorithm, improves the robustness of the model, greatly improves the cloth defect positioning accuracy, meets the cloth defect detection precision, shortens the 3/4 reasoning time, and accelerates the industrial-grade textile mill cloth detecti