A Ship Detection Model Based on YOLOX with Lightweight Adaptive Channel Feature Fusion and Sparse Data Augmentation

Zhang, Quan; Huang, Yanrong; Song, Rui

doi:10.1109/avss56176.2022.9959441

Cited by 4 publications

(7 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…While Section 4 presents a range of trackers that employ the methods discussed here for diverse scenarios such as pedestrians, cars, and objects, the area of maritime surveillance using ship trackers remains an unexplored field. For visual ship detection, we can find recent techniques in [11,43] that utilize state-of-the-art approaches.…”

Section: Workmentioning

confidence: 99%

Beyond Land: A Review of Benchmarking Datasets, Algorithms, and Metrics for Visual-Based Ship Tracking

Rocha

Figueiredo

2023

Electronics

View full text Add to dashboard Cite

Object tracking has gained much interest in the last few years, especially in the context of multiple object tracking. Many datasets used for tracking provide video sequences of people and objects in very different contexts. Although it has been attracting much attention, no dataset or tracking algorithm has been applied to coastal surveillance and ship tracking. Besides image/video-based tracking technologies, other technologies, such as radar and automatic identification systems (AISs), are also used for this task, especially in maritime applications. In the AIS case, commonly known issues, such as information omission, remain to be dealt with. As for radars, the most important issue is the impossibility of identifying the ship type/class and correlating it with AIS information. However, image/video-based solutions can be combined with these technologies to mitigate or even solve these issues. This work aims to review the most recent datasets and state-of-the-art tracking algorithms (also known as trackers) for single or multiple objects tracking for objects in general and its possibilities for maritime scenarios. The goal is to gain insights for developing novel datasets; benchmarking metrics; and mainly, novel ship tracking algorithms.

show abstract

Section: Workmentioning

confidence: 99%

Beyond Land: A Review of Benchmarking Datasets, Algorithms, and Metrics for Visual-Based Ship Tracking

Rocha

Figueiredo

2023

Electronics

View full text Add to dashboard Cite

show abstract

“…Light_SDNet [21] modified the YOLO5 backbone by a Gost Unit [45] and DepthWise Convolution (DWConv) [46] to reduce the number of parameters; also, data augmentations like haze generation and rain generation have been introduced to enrich the training set. Recently, YOLOX has been considered a robust and powerful method for object detection; Zhang_2022 [47] used the YOLOX framework to design a lightweight method. Instead of using a PANnet [48] for feature fusion, the paper used a Lightweight Adaptive Channel Feature Fusion (LACFF) to overcome the inconsistent scale of feature maps.…”

Section: Ship Detectionmentioning

confidence: 99%

“…Afterward, the channels are fused according to the learned weights. Similar to Zhang_2022 [47], our work is also based on YOLOX; however, we do not focus on feature fusion but introduce a loss that selects suitable features on the classification head.…”

Section: Ship Detectionmentioning

confidence: 99%

“…There are several experiment settings from different works. Given 70,000 published images from SeaShip [19], Zhang_2022 [47], and Zhang_2021 [42] used 90% of data for training and validating; the other 10% of data are the testing dataset. Liu_2020 [39], Liu_2022 [36], Han_2021 [44], and Light_SDNet [21] use 80% of data for training and validating dataset, the other 20% is the testing dataset.…”

Section: Compare With Sotamentioning

confidence: 99%

See 1 more Smart Citation

Image-Based Ship Detection Using Deep Variational Information Bottleneck

Ngo,

Vo,

Nguyen

et al. 2023

Sensors

View full text Add to dashboard Cite

Image-based ship detection is a critical function in maritime security. However, lacking high-quality training datasets makes it challenging to train a robust supervision deep learning model. Conventional methods use data augmentation to increase training samples. This approach is not robust because the data augmentation may not present a complex background or occlusion well. This paper proposes to use an information bottleneck and a reparameterization trick to address the challenge. The information bottleneck learns features that focus only on the object and eliminate all backgrounds. It helps to avoid background variance. In addition, the reparameterization introduces uncertainty during the training phase. It helps to learn more robust detectors. Comprehensive experiments show that the proposed method outperforms conventional methods on Seaship datasets, especially when the number of training samples is small. In addition, this paper discusses how to integrate the information bottleneck and the reparameterization into well-known object detection frameworks efficiently.

show abstract

“…Some studies have designed algorithms for maritime conditions and achieved good results [19], such as ISDet, which improved the accuracy of marine target recognition by enhancing the ShuffleNet network structure and applying the PD-NAML training method [20]; CLFR-Det, which enhanced recognition accuracy by using features of different levels and semantics, combined with cross-layer deformed convolution and a multi-scale feature refinement mechanism for enriching the semantic information of small vessels [21]; and methods that improved recognition accuracy by merging multiple visual features and segmenting sea-surface images after detecting the sea-sky line [22]. YOLO-based object detection algorithms, tailored for marine targets, have also demonstrated real-time target recognition capabilities, capable of recognizing ship targets in satellite [23][24][25], aerial [26], and horizontal perspective images [27], proving the feasibility of deep learning-based marine target recognition algorithms. However, they often require data fusion with spatial information obtained from radar to locate targets [28], hindering their use under radio silence conditions for maritime target localization.…”

Section: Introductionmentioning

confidence: 99%

StereoYOLO: A Stereo Vision-Based Method for Maritime Object Recognition and Localization

Shang,

Yu,

Zeng

et al. 2024

JMSE

View full text Add to dashboard Cite

Image recognition is vital for intelligent ships’ autonomous navigation. However, traditional methods often fail to accurately identify maritime objects’ spatial positions, especially under electromagnetic silence. We introduce the StereoYOLO method, an enhanced stereo vision-based object recognition and localization approach that serves autonomous vessels using only image sensors. It is specifically refined for maritime object recognition and localization scenarios through the integration of convolutional and coordinated attention modules. The method uses stereo cameras to identify and locate maritime objects in images and calculate their relative positions using stereo vision algorithms. Experimental results indicate that the StereoYOLO algorithm boosts the mean Average Precision at IoU threshold of 0.5 (mAP50) in object recognition by 5.23%. Furthermore, the variation in range measurement due to target angle changes is reduced by 6.12%. Additionally, upon measuring the distance to targets at varying ranges, the algorithm achieves an average positioning error of 5.73%, meeting the accuracy and robustness criteria for maritime object collision avoidance on experimental platform ships.

show abstract

A Ship Detection Model Based on YOLOX with Lightweight Adaptive Channel Feature Fusion and Sparse Data Augmentation

Cited by 4 publications

References 23 publications

Beyond Land: A Review of Benchmarking Datasets, Algorithms, and Metrics for Visual-Based Ship Tracking

Beyond Land: A Review of Benchmarking Datasets, Algorithms, and Metrics for Visual-Based Ship Tracking

Image-Based Ship Detection Using Deep Variational Information Bottleneck

StereoYOLO: A Stereo Vision-Based Method for Maritime Object Recognition and Localization

Contact Info

Product

Resources

About