Abstract: In order to improve the accuracy and speed of apple leaf disease detection, this paper proposes an improved SSD algorithm to achieve the identification of apple leaf diseases. In order to extract the information of pathologies in leaves more effectively, the base network of SSD was replaced with ResNet50, and ResNet50 was also improved by combining the attention mechanism. The feature fusion structure was used in the SSD network to enhance the algorithm's ability to recognize small target information. To improve the recognition accuracy more effectively, a new feature enhancement module was proposed to enhance the deep mapping of the network. The results demonstrate that the mean average precision of the improved SSD network reaches 91.57%, which is 8.17 percentage points higher than the original SSD model, and the detection speed reaches 50.3 frame/s, effectively improving the accuracy of apple leaf disease detection.