Abstract:Long-term monitoring and recognition of underwater organism objects are of great significance in marine ecology, fisheries science and many other disciplines. Traditional techniques in this field, including manual fishing-based ones and sonar-based ones, are usually flawed. Specifically, the method based on manual fishing is time-consuming and unsuitable for scientific researches, while the sonar-based one, has the defects of low acoustic image accuracy and large echo errors. In recent years, the rapid development of deep learning and its excellent performance in computer vision tasks make vision-based solutions feasible. However, the researches in this area are still relatively insufficient in mainly two aspects. First, to our knowledge, there is still a lack of large-scale datasets of underwater organism images with accurate annotations. Second, in consideration of the limitation on hardware resources of underwater devices, an underwater organism detection algorithm that is both accurate and lightweight enough to be able to infer in real time is still lacking. As an attempt to fill in the aforementioned research gaps to some extent, we established the Multiple Kinds of Underwater Organisms (MKUO) dataset with accurate bounding box annotations of taxonomic information, which consists of 10,043 annotated images, covering eighty-four underwater organism categories. Based on our benchmark dataset, we evaluated a series of existing object detection algorithms to obtain their accuracy and complexity indicators as the baseline for future reference. In addition, we also propose a novel lightweight module, namely Sparse Ghost Module, designed especially for object detection networks. By substituting the standard convolution with our proposed one, the network complexity can be significantly reduced and the inference speed can be greatly improved without obvious detection accuracy loss. To make our results reproducible, the dataset and the source code are available online at https://cslinzhang.github.io/MKUO-and-Sparse-Ghost-Module/ .

SUNOD: Synthetic Underwater Non-Natural Object Detection Dataset

YoloXT: A Object Detection Algorithm for Marine Benthos

Rethinking General Underwater Object Detection: Datasets, Challenges, and Solutions

Underwater Object Detection Based on Enhanced YOLO

Real-time Underwater Object Detection Technology for Complex Underwater Environments Based on Deep Learning

USOD10K: A New Benchmark Dataset for Underwater Salient Object Detection

A Dataset and Benchmark of Underwater Object Detection for Robot Picking

A Benchmark dataset for both underwater image enhancement and underwater object detection

Underwater object detection and datasets: a survey

An Underwater Organism Image Dataset and a Lightweight Module Designed for Object Detection Networks

A YOLOv5 Baseline for Underwater Object Detection

A Dataset with Multibeam Forward-Looking Sonar for Underwater Object Detection

A Dataset for Detection and Segmentation of Underwater Marine Debris in Shallow Waters

YOLOv5-UOD: Underwater Object Detection Method Based on Improved YOLOv5

Underwater Object Detection in the Era of Artificial Intelligence: Current, Challenge, and Future

Multi-modal Perception Dataset of In-water Objects for Autonomous Surface Vehicles

UWStereo: A Large Synthetic Dataset for Underwater Stereo Matching

Semantic Segmentation of Underwater Imagery: Dataset and Benchmark

Self-Supervised Marine Organism Detection from Underwater Images

An Underwater Image Enhancement Benchmark Dataset and Beyond

A New Dataset, Poisson GAN and AquaNet for Underwater Object Grabbing