In the detection of small defects in the glass of mobile phone cover plates, an improved SSD model is proposed to address the issue of missed detection in traditional single shot multi-layer detector (SSD) models. This model uses ResNet50 as the backbone network, while incorporating dilation convolution and attention mechanism SeNet to improve the accuracy of small defect detection. Experimental comparisons were conducted on the defect dataset, and the results showed that the improved SSD model had higher detection accuracy and lower miss rate compared to traditional SSD models, which could meet the accuracy requirements of defect detection.