ROIMIX: Proposal-Fusion Among Multiple Images for Underwater Object Detection

Lin, Weihong; Zhong, Jia-Xing; Liu, Shan; Li, Thomas; Li, Ge

doi:10.1109/icassp40776.2020.9053829

“…A detector termed Domain Generalization YOLO (DG-YOLO) is then provided to extract semantic information from WQT data, where YOLOv3, Domain Invariant Module, and Invariant Risk Minimization penalty are three basic components. In [92], an approach called RoIMix is used to create different training examples by fusing several pictures at the proposal level. The RoIMix seeks to mimic overlapping, occluded, and blurred items so that the model may learn to recognize underwater organisms implicitly.…”

Section: Data Augmentationmentioning

confidence: 99%

Research Challenges, Recent Advances and Benchmark Datasets in Deep-Learning-Based Underwater Marine Object Detection: A Review

Er¹,

Chen²

2022

Preprint

0

View full text Add to dashboard Cite

Underwater marine object detection, as one of the most fundamental techniques in marine engineering, had been shown to exhibit tremendous potential for exploring the oceans in recent years. It has been applied widely in monitoring of underwater ecosystems, exploration of natural resources, management of commercial fishery, and so on. However, due to the complexity of the underwater environment, characteristics of marine objects, and limitations from exploration equipment, the detection performance in terms of speed, accuracy and robustness could be degraded dramatically using conventional approaches. Deep learning algorithms have found significant impact in many fields of engineering and science including marine engineering. In this context, We offer an overview of deep-learning-based underwater marine object detection techniques. We organize the present research challenges into three areas to assist a thorough knowledge of the subject matter, namely image quality degradation, small object detection, and poor generalization. We evaluate recent advances and emphasize the advantages and risks of existing solutions for each category of extant challenges. In addition, we examine the most extensively used benchmark datasets for underwater marine object detection in detail and critically. Comparisons between previous reviews and prospective advancements in the subject field, notably Artificial-Intelligence-based approaches, are also presented.

show abstract

“…In the year of 2017, underwater object detection for open-sea farming is first proposed in the target recognition track of Un- 2 MMDetection is an open source object detection toolbox based on Py-Torch. https://github.com/open-mmlab/mmdetection 3 JETSON AGX XAVIER is an embedded development board produced by NVIDIA which could be deployed in an underwater robot.…”

Section: Introductionmentioning

confidence: 99%

A Dataset and Benchmark of Underwater Object Detection for Robot Picking

Liu¹,

Li

²

,

Wang³

et al. 2021

2021 IEEE International Conference on Multimedia &Amp; Expo Workshops (ICMEW)

View full text Add to dashboard Cite

Underwater object detection for robot picking has attracted a lot of interest. However, it is still an unsolved problem due to several challenges. We take steps towards making it more realistic by addressing the following challenges. Firstly, the currently available datasets basically lack the test set annotations, causing researchers must compare their method with other SOTAs on a self-divided test set (from the training set). Training other methods lead to an increase in workload and different researchers divide different datasets, resulting there is no unified benchmark to compare the performance of different algorithms. Secondly, these datasets also have other shortcomings, e.g., too many similar images or incomplete labels. Towards these challenges we introduce a dataset, Detecting Underwater Objects (DUO), and a corresponding benchmark, based on the collection and re-annotation of all relevant datasets. DUO contains a collection of diverse underwater images with more rational annotations. The corresponding benchmark provides indicators of both efficiency and accuracy of SOTAs (under the MMDtection framework) for academic research and industrial applications, where JET-SON AGX XAVIER is used to assess detector speed to simulate the robot-embedded environment.

show abstract

“…Unfortunately, as the information listed in Table 1, URPC series datasets do not provide the annotation file of the test set and cannot be downloaded after the contest. Therefore, researchers [1,2] first have to divide the training data into two subsets, including a new subset of training data and a new subset of testing data, and then train their proposed method and other SOTA methods. On the one hand, training other methods results in a significant increase in workload.…”

Section: Introductionmentioning

confidence: 99%

“…• We provide a corresponding benchmark of SOTA detectors on DUO including efficiency and accuracy indicators which could be a reference for both academic research and industrial applications. 2 MMDetection is an open source object detection toolbox based on Py-Torch. https://github.com/open-mmlab/mmdetection 3 JETSON AGX XAVIER is an embedded development board produced by NVIDIA which could be deployed in an underwater robot.…”

Section: Introductionmentioning

confidence: 99%

A Dataset And Benchmark Of Underwater Object Detection For Robot Picking

Liu,

Li,

Wang

et al. 2021

Preprint

View full text Add to dashboard Cite

Underwater object detection for robot picking has attracted a lot of interest. However, it is still an unsolved problem due to several challenges. We take steps towards making it more realistic by addressing the following challenges. Firstly, the currently available datasets basically lack the test set annotations, causing researchers must compare their method with other SOTAs on a self-divided test set (from the training set). Training other methods lead to an increase in workload and different researchers divide different datasets, resulting there is no unified benchmark to compare the performance of different algorithms. Secondly, these datasets also have other shortcomings, e.g., too many similar images or incomplete labels. Towards these challenges we introduce a dataset, Detecting Underwater Objects (DUO), and a corresponding benchmark, based on the collection and re-annotation of all relevant datasets. DUO contains a collection of diverse underwater images with more rational annotations. The corresponding benchmark provides indicators of both efficiency and accuracy of SOTAs (under the MMDtection framework) for academic research and industrial applications, where JET-SON AGX XAVIER is used to assess detector speed to simulate the robot-embedded environment.

show abstract

ROIMIX: Proposal-Fusion Among Multiple Images for Underwater Object Detection

Cited by 95 publications

References 28 publications

Research Challenges, Recent Advances and Benchmark Datasets in Deep-Learning-Based Underwater Marine Object Detection: A Review

Research Challenges, Recent Advances and Benchmark Datasets in Deep-Learning-Based Underwater Marine Object Detection: A Review

A Dataset and Benchmark of Underwater Object Detection for Robot Picking

A Dataset And Benchmark Of Underwater Object Detection For Robot Picking

Contact Info

Product

Resources

About