Scholarlink

Underwater trash detection plays an important area of marine conservation, undoing the environmental effects of pollution. This study compares the performance of four deep learning architectures—R-CNN, YOLO, SSD, and DeepLabv3+ —for object detection and classification of underwater trash. The models were tested for accuracy, precision, recall, and F1-score to determine their real-world applications. DeepLabv3+ performed the best with 94.34% accuracy and an F1-score of 91.16%, with excellent performance in pixel-wise segmentation and small object detection in challenging underwater environments. While YOLO and SSD offer speed and efficiency for real-time applications, R-CNN offers accurate object localization but suffer from occlusions. The results enhance automatic marine trash detection, paving the way for real-time applications in underwater surveillance and cleanup operations.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Comparative Analysis of Deep Learning Models for Underwater Trash Detection: A Study of R-CNN, YOLO, SSD, and DeepLabv3+

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Kaustubh Patil,
  • Vaidehi Tiwari,
  • Anjana Rajagopal,
  • Rachana Nair,
  • Naman Chachan,
  • Mamta Bhamare

摘要

Underwater trash detection plays an important area of marine conservation, undoing the environmental effects of pollution. This study compares the performance of four deep learning architectures—R-CNN, YOLO, SSD, and DeepLabv3+ —for object detection and classification of underwater trash. The models were tested for accuracy, precision, recall, and F1-score to determine their real-world applications. DeepLabv3+ performed the best with 94.34% accuracy and an F1-score of 91.16%, with excellent performance in pixel-wise segmentation and small object detection in challenging underwater environments. While YOLO and SSD offer speed and efficiency for real-time applications, R-CNN offers accurate object localization but suffer from occlusions. The results enhance automatic marine trash detection, paving the way for real-time applications in underwater surveillance and cleanup operations.