Lightweight apple target detection method

The invention discloses a lightweight apple target detection method, and relates to the technical field of apple target detection, and the method specifically comprises the following steps: S1, providing a VanillaC2f structure for a backbone network; s2, for the neck network, an OfficientNeck structure is provided; and S3, for the loss function of the detection frame, replacing the original CIoU bounding box loss function with the WIoU bounding box loss function of the dynamically updated gradient gain distribution strategy. According to the method, the original YOLOv8n backbone network is replaced by the VanilaNet with low complexity, so that the parameter quantity and the calculation quantity of the backbone network are reduced, and the lightweight of the backbone network is realized; at the same time, the neck network is improved by using efficient OfficientNeck, and the lightweight of the neck network is realized by using a ghosting shuffling module and a one-time aggregation cross-stage local network mod