In order to solve the problems of low accuracy, missed detection and false detection of stacked objects in the existing object detection, an improved algorithm model based on YOLOV8 was proposed. The model introduces Deformable Convolutiona Networks. The Shueffle Attention mechanism is added to reduce the complexity of the model and improve the ability to express features. The Dysample upsampling module is introduced to replace the original Upsample module, which reduces the requirement of computing resources. The improved model is trained on the stacked object dataset, and compared with the original model, the accuracy of the improved network model is improved by 1.7%, and the accuracy of mAP50-95 is increased by 1%. It provides a theoretical reference for the study of object detection of stacked objects.