<![CDATA[This study developed an object detection model using YOLOv9 to identify pig feces on pig skin, addressing challenges in automating pig cleaning systems and reducing the spread of African Swine Fever (ASF). The aim was to enhance biosecurity measures by minimizing human-pig contact through automation. A specialized dataset comprising 5,404 images was collected from Nyoman Farm in Bali, Indonesia, under various lighting and cleanliness conditions. These images were annotated into two classes, namely 'feces' and 'pig,' following strict criteria to ensure clarity and distinction. YOLOv9 was chosen as it is an advanced update of YOLOv8 with enhanced object detection capabilities. The model was iteratively trained and optimized to achieve the best performance. The results achieved a mAP_0.5 of 70.5%, precision of 70.6%, and recall of 72.1%. However, the model faced challenges in distinguishing pig skin patterns from feces and managing false positives caused by similar-looking objects in the barn environment. Despite these challenges, integrating this model into an automated cleaning system can reduce human-pig contact by up to 76%, which is expected to significantly lower the risk of ASF transmission. This study contributes to automated farming technology, demonstrating how well YOLOv9 can detect complex objects in agricultural settings and providing practical solutions to enhance biosecurity in pig farming while improving productivity.]]>
Copyrights © 2025
