Abstract:
Seeds circulating in Indonesia are required to have seed eligibility certificates
regulated by PERMENTAN No. 12 of 2018. One of the processes to obtain seed
feasibility certification is the purity analysis laboratory test. In this process, the seed
analyst sorts working samples of seeds that contain several components, including pure
seeds and non-pure seeds. The sorting is done manually by identifying the seeds in the
working samples based on their morphological appearance. In this study, object
detection of rice seeds was conducted using the "You Only Look Once (YOLO)"
algorithm. YOLO is a real-time object detection algorithm. The method used is transfer
learning with the pre-trained YOLOv5s model, which is one of the models from YOLO
version 5. The dataset consists of 235 photos of working examples of rice seeds, with
two classes labeled as rice and non-rice. The evaluation results of the overall model
performance are as follows: precision of 0.908, recall of 0.808, mAP@0.5 of 0.859, and
F1_Score of 0.85.
