From Pixels to Plastic: Charting a Path to Detect, Track, and Count Oceanic Plastic Pollution through Computer Vision

27 Jul 2023 (modified: 07 Dec 2023)DeepLearningIndaba 2023 Conference SubmissionEveryoneRevisionsBibTeX
Keywords: Plastic pollution in Oceans, Computer Vision, Object Detection, Object Tracking, Object counting
TL;DR: Detect, track, and count oceanic plastic pollution through computer vision
Abstract: Plastic pollution poses an alarming threat to marine ecosystems, necessitating innovative and efficient solutions for its monitoring and management. Building upon recent advancements in Artificial Intelligence (AI), Machine Learning (ML), and Deep Learning (DL) [7], we have developed a cuttingedge, AI-driven model utilizing the YOLOv8 architecture[4]. This model not only excels in real-time detection of marine plastics but also integrates advanced tracking and counting capabilities. Tailored for compatibility with marine robotics and other low-resource applications, our approach offers a robust solution even in GPU-deprived environments setting it apart from previous efforts employing the R-CNN architecture[3]. Other recent studies have employed imaging technologies coupled with deep learning techniques, such as the deployment of bridge-mounted cameras on rivers in Jakarta and the use of Unmanned Aerial Systems (UAS) for marine litter mapping on beach-dune systems. Our model notably surpasses these methodologies in performance, achieving superior precision, recall, and overall efficiency metrics. Beyond its detection prowess, our model represents a paradigm shift in the computational efficiency of monitoring tools, poised to revolutionize the strategies to combat the plastic pollution menace in aquatic ecosystems. Keywords:Plastic pollution in Oceans,Computer Vision,Object Detection,Object Tracking,Object counting
Submission Category: Machine learning algorithms
Submission Number: 37
Loading