TRAINING METHOD AND APPARATUS FOR CLASSIFICATION NEURAL NETWORK FOR SEMANTIC SEGMENTATION, AND ELECTRONIC DEVICE

摘要

A training method and apparatus for a classification neural network for semantic segmentation, and an electronic device. By setting the weights of uncertain pixels in a boundary region to be small during the calculation of the loss of the neural network, the impact of these pixels on the training speed and model stability can be reduced, thereby effectively increasing the training speed and model stability, and the accuracy of the training result of the main part in a training image can be ensured.