Fabric defect detection method and system based on neural network

A fabric defect detection method and a fabric defect detection system based on neural network are provided. The fabric defect detection method includes following steps. In a training phase: A plurality of defect images are generated. The plurality of generated defect images and at least one positive sample fabric image are synthesized to obtain a plurality of negative sample fabric images. A neural network model is trained according to the plurality of generated defect images, the plurality of negative sample fabric images and a loss function. In an inference phase: A fabric image is captured. The fabric image is detected by the neural network model, and a fabric defect image is output. The fabric defect image is a binary image.