Deep Learning Technique Detection for Cotton and Leaf Classification Using the YOLO Algorithm
Julie Ann B. Susa, Wendy C. Nombrefia, Alfredo S. Abustan, Jonel R. Macalisang, Renato R. Maaliw
Abstract
Cotton is one of the world's most significant crops, and it is widely grown. It is vulnerable to a number of plant diseases, resulting in a significant reduction in yield and productivity. It is critical to discover a disease at an early stage to receive prompt diagnosis and treatment. As a result, a deep learning approach was used to suggest a cotton plant classification system. The YOLOv3 algorithm was used in the study, which is the most important real-time object identification system for detecting and classifying damaged and healthy plants and leaves. The applied model has an mAP (mean Average Precision) score of 96.09 %, with training accuracy of 96.79 % and validation accuracy of 92.26 %. The detection accuracy of video frames ranges between 98 and 99 % in the testing results, whereas the detection accuracy of live stream image frames ranges between 74 and 99 %. As a result, the model outperformed other current algorithms and is the best choice for cotton plant detection and categorization.