Melnychenko, OleksandrSavenko, OlegRadiuk, Pavlo2024-01-082024-01-082023-12-21Melnychenko O., Savenko O., Radiuk P. Apple detection with occlusions using modified YOLOv5-v1. The 12th IEEE International Conference on Intelligent Data Acquisition and Advanced Computing Systems: Technology and Applications (IDAACS’2023) : Proceedings. Vol. 1. (Dortmund, Germany, 7-9 September 2023). IEEE Inc., 2023. P. 107-112. (Scopus, Q4). DOI: https://doi.org/10.1109/IDAACS58523.2023.103487792770-4254https://elar.khmnu.edu.ua/handle/123456789/15236In our research, we created a novel YOLOv5-v1 architecture to identify apples in images with occlusions. We specifically engineered new layers for the BottleneckCSP-v4 module, which replaces the original BottleneckCSP module within the backbone structure of the YOLOv5 network. Integrating the SENet module into our improved trunk network helps to discern features of medium and large-sized fruits more accurately under varying conditions. We also adjusted the initial size of the binding block within the source network to avoid incorrect identification of small objects within the image's background. Based on the test dataset, our experimental results show that our advanced network model can effectively identify fruits captured through an unmanned aerial vehicle camera. The classification metrics - recall, precision, mAP, and F1-score - obtained scores of 92.13%, 84.59%, 87.94%, and 89.02% respectively.enimage processingobject detectionapple yieldYOLOv5deep learningvisual occlusionsТези доповідей