2021/08/15 by Deepak Singh, Singh, Deepak, Matías Valdenegro-Toro +1 · 2 citations
Computer Science · Environmental Science · #Advanced Neural Network Applications #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Robotics (cs.RO) #Water Quality Monitoring Technologies
paper · pdf · doi:10.48550/arxiv.2108.06800
openalex publication_date 2021/08/15 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Accurate detection and segmentation of marine debris is important for keeping\nthe water bodies clean. This paper presents a novel dataset for marine debris\nsegmentation collected using a Forward Looking Sonar (FLS). The dataset\nconsists of 1868 FLS images captured using ARIS Explorer 3000 sensor. The\nobjects used to produce this dataset contain typical house-hold marine debris\nand distractor marine objects (tires, hooks, valves,etc), divided in 11 classes\nplus a background class. Performance of state of the art semantic segmentation\narchitectures with a variety of encoders have been analyzed on this dataset and\npresented as baseline results. Since the images are grayscale, no pretrained\nweights have been used. Comparisons are made using Intersection over Union\n(IoU). The best performing model is Unet with ResNet34 backbone at 0.7481 mIoU.\nThe dataset is available at\nhttps://github.com/mvaldenegro/marine-debris-fls-datasets/\n