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

Liu, Chongwei; Wang, Zhihui; Wang, Shijie; Tang, Tao; Tao, Yulong; Yang, Caifei; Li, Haojie; Liu, Xing; Fan, Xin

doi:10.1109/tcsvt.2021.3100059

Cited by 40 publications

(35 citation statements)

References 47 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [8], an efficient object detection network (AquaNet) based on CenterNet [38] was presented, in which multi-scale feature maps from different phases are fused by the FPN for recognizing mass tiny objects from foggy underwater photos. In [39], the Multi-scale ResNet (M-ResNet) based on a modified residual neural network for underwater object detection is proposed that leverages multi-scale feature fusing by the FPN to allow accurate detection of objects of varied sizes, especially small ones.…”

Section: Multi-scale Representationmentioning

confidence: 99%

“…The process of AFFM can be described as follows: Different from the conventional feature fusion that cross multi layers in the backbone, the idea that fuse multi-scale features that cross different channels using different sizes of kernels or complex connections in one block is proposed in [50,51]. A depth-wise-convolution-based Multi-scale Contextual Features Fusion (MFF) block is proposed in AquaNet [8], which can not only represents multi-scale features at a granular level and but also expands the range of receptive fields for each network layer. As shown in Figure 20, for a given input, the number of channels is first expanded N times (N is a sequence where the elements according to the kernel size described as follows, e.g., [3,5,7]) by 1 × 1 convolution.…”

Section: Multi-scale Representationmentioning

confidence: 99%

“…Underwater object detection can be applied in providing important information for the cultivation, status surveillance and early warning of diseases [7]. Underwater object detection can also act as a foundation for the robot grabbing tasks, e.g., picking holothurian, sea urchin, scallop, and other marine products [8]. These shreds of evidence show that the importance of underwater object detection cannot be overemphasized.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

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

Er¹,

Chen²

2022

Preprint

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

Section: Multi-scale Representationmentioning

confidence: 99%

Section: Multi-scale Representationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

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

Er¹,

Chen²

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…In [11], an efficient object detection network (AquaNet) based on CenterNet [53] was presented, in which multiscale feature maps from different phases are fused by the FPN for recognizing mass tiny objects from foggy underwater photos. In [54], a modified residual neural network for underwater object detection termed Multi-scale ResNet (M-ResNet) was proposed, which uses multi-scale feature fusion by the FPN, allowing accurate detection of objects of various sizes, especially small objects.…”

Section: Multi-scale Representationmentioning

confidence: 99%

“…It can be applied in marine ranching, providing important information for the cultivation, status surveillance and early warning of diseases [10]. On the other hand, underwater object detection can act as a foundation for the robot's grabbing tasks, e.g., picking holothurian, sea urchin, scallop, and other marine products [11]. These shreds of evidence show that the importance of underwater object detection cannot be overemphasized.…”

Section: Introductionmentioning

confidence: 99%

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

Er¹,

Chen²

2022

Preprint

View full text Add to dashboard Cite

Underwater marine object detection, as one of the most fundamental techniques in marine engineering, has been shown to exhibit significant potential for exploring underwater environment in recent years. It has been applied widely in the monitoring of underwater ecosystems, exploration of natural resources, management of commercial fishery management, and other areas. However, due to the complexity of underwater environment, characteristics of marine object, and limitation from exploration equipment, the detection performance revolving around speed, accuracy and robustness could be degraded dramatically. In this context, we present a survey of deep-learning-based underwater marine object detection. To facilitate comprehensive understanding of the subject matter, we categorize the existing research challenges of underwater object detection into image quality degradation, small object detection, poor generalization and real-time detection. Corresponding to each category of the existing challenges, we review recent advances and highlight the pros and cons of existing techniques. Furthermore, we give a detailed and critical review of the most widely used benchmark datasets for underwater marine object detection. Comparisons with existing reviews and future trends of the subject matter, particularly AI-based techniques, are also presented.

show abstract

Brain Lesion Synthesis via Progressive Adversarial Variational Auto-Encoder

Huo

Vakharia

et al. 2022

Simulation and Synthesis in Medical Imaging

View full text Add to dashboard Cite

Accurate brain lesion delineation is important for planning neurosurgical treatment. Automatic brain lesion segmentation methods based on convolutional neural networks have demonstrated remarkable performance. However, neural network performance is constrained by the lack of large-scale well-annotated training datasets. In this manuscript, we propose a comprehensive framework to efficiently generate new, realistic samples for training a brain lesion segmentation model. We first train a lesion generator, based on an adversarial autoencoder, in a selfsupervised manner. Next, we utilize a novel image composition algorithm, Soft Poisson Blending, to seamlessly combine synthetic lesions and brain images to obtain training samples. Finally, to effectively train the brain lesion segmentation model with augmented images we introduce a new prototype consistence regularization to align real and synthetic features. Our framework is validated by extensive experiments on two public brain lesion segmentation datasets: ATLAS v2.0 and Shift MS. Our method outperforms existing brain image data augmentation schemes. For instance, our method improves the Dice from 50.36% to 60.23% compared to the U-Net with conventional data augmentation techniques for the ATLAS v2.0 dataset.

show abstract

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

Cited by 40 publications

References 47 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

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

Brain Lesion Synthesis via Progressive Adversarial Variational Auto-Encoder

Contact Info

Product

Resources

About