Re-optimization training method based on existing image semantic segmentation model and application

The invention discloses a re-optimization training method based on an existing image semantic segmentation model and application. The method includes the steps of outputting the last layer of an imagesemantic segmentation neural network model; and intercepting a plurality of prediction tags with the highest prediction probability from all pixels close to a semantic edge, performing feature distance measurement and calculation through a re-optimization model, and taking the nearest tag as the correction prediction tag of the pixel, thereby achieving the purpose of improving the semantic segmentation prediction accuracy. The invention provides a boundary deviation elimination method based on re-identification, eliminates uncertainty of a semantic edge adjacent region, and is an improvementof a mature image semantic segmentation model. The optimization model focuses on the correction task of the semantic edge. In addition, only the image semantic edge area is optimized, and excessive operation time and space bur