An Improved Soft-CBAM-YoloV5 Algorithm for Fruits and Vegetables Detection and Counting

The quality of people's life has been improved as science and technology keep progressing, and artificial intelligence has been implemented into more and more fields these days. Among them, many researchers have paid attention to the scenes of quoting prices and settling accounts for fruits and vegetables in supermarkets and manual classification of fruits and vegetables in farmers' markets. In the traditional YOLOV5 method, when the same kind of fruits and vegetables are blocked from each other, the counting will be inaccurate, resulting in missed detection. In addition, the environment where different objects stay will affect the feature extraction of objects. In order to reduce the occurrence of missed detections as much as possible and improve the detection performance of the model, this paper proposes a new detection model counting method named Soft-CBAM-YOLOV5, by reasonably reducing the score of adjacent detection boxes and introducing the attention mechanism CBAM (Convolutional Block Attention Module) to prevent the occurrence of missed detection and strengthen the focus on the detection target to improve the feature extraction ability of the algorithm. Compared with traditional YOLOV5, the mAP value of Soft-CBAM-YOLOV5 increases from 97.05% to 97.71%. Consequently, the Soft-CBAM-YOLOV5 model has greater detection performance, robustness and lower missed detection rate.