Surface defect image segmentation method based on weight fusion and double encoders

The invention discloses a surface defect image segmentation method based on weight fusion and double encoders, and the method comprises the following steps: constructing a surface defect image semantic segmentation network which comprises a feature extraction module, a feature fusion module and a double-encoding module; inputting the image set into a feature extraction module to obtain feature maps of different levels; inputting the feature maps of different levels into a feature fusion module to obtain enhanced feature maps of each level; inputting the enhanced feature maps of each level into a dual-coding module to obtain segmented maps of different levels; calculating weighted cross entropy loss and weighted intersection-to-union ratio loss, and performing weighted summation to obtain total loss; using total loss function constraint to complete training of a surface defect image semantic segmentation network, and performing a surface defect image segmentation task; according to the method, the receptive fi