Research on detection and location of tomato axillary bud growth point via lightweight YOLOv3 model

The removal of tomato axillary buds is of great significance for improving fruit yield and quality. In response to the current lack of effective identification and localization of tomato axillary bud growth points, a growth point detection and localization method based on lightweight YOLOv3 is proposed. Improve the traditional YOLOv3 model architecture and introduce the lightweight recognition model MobileNet to replace the YOLOv3 feature extraction network, ensuring detection accuracy while improving detection speed. Replacing some standard convolutions with deep separable convolutions reduces model parameters without compromising network performance. The comparative experimental results show that the average recognition accuracy of the improved YOLOv3 model has been effectively improved, reducing the detection time of each image, and reducing the number of parameters and weight files compared to the original model. This will help to achieve detection and localization of tomato axillary bud growth points on airborne low-performance terminals in the future.