This paper presents an improved layered clothing detection algorithm based on improved YOLOV9(You Only Look Once), marking the first algorithm specifically designed for layered clothing recognition. Specifically, the SE attention mechanism was added to the YOLOV9 backbone neural network to improve the utilization of clothing features during the training process. They are changing the loss function of YOLOV9 to GIoU (Generalized Intersection over Union) to improve the recognition accuracy under different postures and the recognition effect when key features are occluded. At the same time, this paper produces and uses a new Layered_Deepfashion2 dataset to improve the recognition effect of the trained algorithm on layered clothing. Compared with the YOLOV9 model, the enhanced mAP_0 model is increased by 4.7%, mAP0.5:0.95 by 5.07%, and the F1-Measure indicator is increased by 1.81%. This model has a good recognition effect and high accuracy for layered clothing and can be applied well to fashion clothing.