Detection of Marine Animals in a New Underwater Dataset with Varying Visibility
In IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2019
Abstract
The increasing demand for marine monitoring calls for robust automated systems to support researchers in gathering information from marine ecosystems. This includes computer vision based marine organism detection and species classification systems. Current state-of-the-art marine vision systems are based on CNNs, which in nature require a relatively large amount of varied training data. In this paper we present a new publicly available underwater dataset with annotated image sequences of fish, crabs, and starfish captured in brackish water with varying visibility. The dataset is called the Brackish Dataset and it is the first part of a planned long term monitoring of the marine species visiting the strait where the cameras are permanently mounted. To the best of our knowledge, this is the first annotated underwater image dataset captured in temperate brackish waters. In order to obtain a baseline performance for future reference, the YOLOv2 and YOLOv3 CNNs were fine-tuned and tested on the Brackish Dataset.
Info
Conference Paper, 2019
In IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2019
UN SDG Classification
DK Main Research Area
Science/Technology
