Cucumber Flower Detection Based on YOLOv5s-SE7 Within Greenhouse Environments

In the process of cucumber cultivation, quantity and appearance time flowers are important factors influencing final yields. However, it is labor-intensive to timely record count flowers. An optional method identify automatically using computer vision technologies based on images taken by cameras installed in greenhouses. there severe problems with a large field view greenhouse environment, that is, foreground-background imbalance, which renders significant gaps between detection accuracy application requirements even for state-of-the-art models like Faster-RCNN, SSD YOLO. This problem can be improved providing specific datasets suitable models. Hence, this paper, two flower wide medium greenhouses constructed. Four attention mechanisms: SE, CA, CBAM SimAM, compared incorporated into YOLOv5s algorithm improve performance during growing states greenhouse. The results indicated our model SE mechanism reached highest recognition rate than other three methods. [email protected] value YOLOv5s-SE7 0.905, was 3.5% higher benchmark YOLOv5s. Meanwhile, outperformed methods such as Faster-RCNN SSD. classification stages, namely bud, bloom faded flower, high 0.847 mAP, suggesting proposed had good effect application.