To enhance the accuracy and efficiency of fish surface pathology recognition, and to address the current issues of heavy reliance on manual processes and low recognition accuracy, this study constructed a fish surface pathology dataset for four types of diseases that are both prevalent and harmful to fish. Based on this, an improved and optimized Resnet-18 model was developed by integrating both spatial attention and SE channel attention mechanisms to create a high-accuracy DBA_Resnet-18 model. A real-time intelligent fish disease recognition and visualization system was also developed using this model. The improved model incorporates the SE channel attention module in the middle of the network and introduces spatial attention at the end of the network.Test results show that the DBA_Resnet-18 model achieved a classification accuracy of 96.75% for fish surface pathologies, which is respectively 1.71, 2.12, 2.37, 2.83, 2.51, 2.23, 2.50, and 3.53 percentage points higher than the commonly used models Resnet-18, Resnet-34, Resnet-50, Resnet-101, Swin Transformer, VGG-16, VGG-19, and AlexNet. The results indicate that the proposed model and the intelligent fish disease recognition visualization system can rapidly and accurately classify different types of fish surface pathologies, achieving an intelligent fish disease recognition system that can be applied in practical environments for diagnosing fish surface pathology types.