We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
YOLO-CE for Accurate Livestock Detection in Challenging Landfill Environments
Summary
This study developed SIDETA, a deep learning-based livestock detection system using an improved YOLO algorithm (YOLO-CE) designed for cluttered landfill environments where livestock grazing poses microplastic contamination risks. The system achieved a precision of 0.812 and mAP@50 of 0.654 on 577 annotated landfill images, outperforming standard YOLO models in complex detection scenarios.
Livestock grazing at landfill sites in Indonesia poses significant public health risks due to microplastic contamination, a persistent issue despite regulations.This study introduces SIDETA defined as "Smart Intelligent Detection and Tracking Application", a Livestock Detection and Counting System powered by YOLO-CE (You Only Look Once -Cluttered Environments), an improved YOLO algorithm designed for complex landfill environments.YOLO-CE incorporates Cross-Stage Partial connections, a Path Aggregation Network, and Coordinate Attention mechanisms to enhance detection accuracy in cluttered and noisy settings.Developed using Python and the Streamlit framework, SIDETA offers a user-friendly platform for livestock monitoring and data documentation.A dataset of 577 manually annotated images from diverse landfill sites was used, partitioned into training (404 images, 70%), validation (115 images, 20%), and test (58 images, 10%) sets.Experimental results show YOLO-CE outperforms traditional YOLO models, achieving a Precision of 0.812, Recall of 0.529, and mAP@50 of 0.654, demonstrating its robustness and reliability.By addressing challenges such as overlapping objects and occlusions, SIDETA bridges cutting-edge deep learning technologies with real-world applications, providing an accurate and scalable solution to mitigate health risks associated with livestock grazing at landfill sites.