error correctionAlthough the deep learning technology has become more and more mature the performance of the deep learning net-work depends on pre-labeled datasets The bias in the training data set su

Although deep learning technology has become increasingly mature, the performance of the deep learning network depends on pre-labeled datasets. Biases in the training dataset, such as noise and imbalance issues, can result in varying weights and accuracy across different training data. Therefore, to validate the stability of the network proposed in this paper, several experiments were conducted using constructed dataset #1 and dataset #2. The number of images in the training set and test set for each experiment was the same, while the images in each training set and test set were different. The final test results are presented in Figure 8 and Figure 9. The experimental results demonstrate that our network achieved good results and exhibited strong stability in these experiments