HighlightNet: Highlighting Low-Light Potential Features for Real-Time UAV Tracking

Fu, Changhong; Dong, Haolin; Ye, Junjie; Zheng, Guangze; Li, Sihang; Zhao, Jilin

doi:10.1109/iros47612.2022.9981070

Cited by 12 publications

(4 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Despite considerable progress, there is a lack of in-depth research on night-time target tracking. In [7] and, [32], lowlight enhancers related to tracking were developed to preprocess data in tracking channels. However, there is weak cooperation between the model and the tracking model, and it is difficult to learn a cascade structure to fill the gap at the feature level.…”

Section: Related Work a Single Object Tracking(sot)mentioning

confidence: 99%

Unsupervised Nighttime Object Tracking Based on Transformer and Domain Adaptation Fusion Network

Wei,

Fu,

Wang

et al. 2024

IEEE Access

View full text Add to dashboard Cite

Section: Related Work a Single Object Tracking(sot)mentioning

confidence: 99%

Unsupervised Nighttime Object Tracking Based on Transformer and Domain Adaptation Fusion Network

Wei,

Fu,

Wang

et al. 2024

IEEE Access

View full text Add to dashboard Cite

“…Variations in brightness [29], contrast [30], gamma correction [31], or the addition of random noise, such as Gaussian or speckle noise [32], to simulate sensor noise or specific lighting conditions are all examples of light noise. These enhancements can be used individually or in tandem to create more complex and diverse lighting conditions.…”

Section: B Light Signal Enhancementmentioning

confidence: 99%

From Augmentation to Inpainting: Improving Visual SLAM With Signal Enhancement Techniques and GAN-Based Image Inpainting

Theodorou,

Velisavljevic,

Dyo

et al. 2024

IEEE Access

View full text Add to dashboard Cite

This paper undertakes a comprehensive investigation that surpasses the conventional examination of signal enhancement techniques and their effects on visual Simultaneous Localization and Mapping (vSLAM) performance across diverse scenarios. Going beyond the conventional scope, the study extends its focus towards the seamless integration of signal enhancement techniques, aiming to achieve a substantial enhancement in the overall vSLAM performance. The research not only delves into the assessment of existing methods but also actively contributes to the field by proposing innovative denoising techniques that can play a pivotal role in refining the accuracy and reliability of vSLAM systems. This multifaceted approach encompasses a thorough exploration of the intricate relationships between signal enhancement, denoising strategies, their cumulative impact on the performance of vSLAM in real-world applications and the innovative use of Generative Adversarial Networks (GANs) for image inpainting. The GANs effectively fill in missing spaces following object detection and removal, presenting a novel state-of-theart approach that significantly enhances overall accuracy and execution speed of vSLAM. This paper aims to contribute to the advancement of vSLAM algorithms in real-world scenarios, demonstrating improved accuracy, robustness, and computational efficiency through the amalgamation of signal enhancement and advanced denoising techniques.INDEX TERMS signal enhancement, denoising techniques, visual SLAM, object detection, simultaneous localization and mapping(SLAM), object detection and generative adversarial network (GAN).

show abstract

“…As UAV tracking becomes more and more widely used, many studies [9]- [11], [28], [29] on UAV object tracking in nighttime environments improve the UAV tracking performance. DarkLighter [9] enhances low-light images to improve UAV tracking performance in nighttime scenes by utilizing unsupervised training while accounting for noise.…”

Section: B Low-light Uav Trackingmentioning

confidence: 99%

“…To achieve semantic-level low-light enhancement, J. Ye et al [10] construct a spatial-channel Transformer-based lowlight enhancer. HighlightNet [11] adapts to illumination variation and excavates the potential object for low-light UAV tracking. Although most low-light UAV tracking methods employ relatively straightforward strategies, they are all extra additions to the basic UAV tracking structure.…”

Section: B Low-light Uav Trackingmentioning

confidence: 99%

Exploring Knowledge Transfer and Distillation in Deep Learning

Li¹

View full text Add to dashboard Cite

Nighttime unmanned aerial vehicle (UAV) tracking has been facilitated with indispensable plug-and-play low-light enhancers. However, the introduction of low-light enhancers increases the extra computational burden for the UAV, significantly hindering the development of real-time UAV applications. Meanwhile, these state-of-the-art (SOTA) enhancers lack tight coupling with the advanced daytime UAV tracking approach. To solve the above issues, this work proposes a novel mutuallearning knowledge distillation framework for nighttime UAV tracking, i.e., MLKD. This framework is constructed to learn a compact and fast nighttime tracker via knowledge transferring from the teacher and knowledge sharing among various students. Specifically, an advanced teacher based on a SOTA enhancer and a superior tracking backbone is adopted for guiding the student based only on the tight coupling-aware tracking backbone to directly extract nighttime object features. To address the biased learning of a single student, diverse lightweight students with different distillation methods are constructed to focus on various aspects of the teacher's knowledge. Moreover, an innovative mutual-learning room is designed to elect the superior student candidate to assist the remaining students frame-by-frame in the training phase. Furthermore, the final best student, i.e., MLKD-Track, is selected through the testing dataset. Extensive experiments demonstrate the effectiveness and superiority of MLKD and MLKD-Track. The practicality of the MLKD-Track is verified in real-world tests with different challenging situations. The code is available at https://github.com/lyfeng001/MLKD.

show abstract

HighlightNet: Highlighting Low-Light Potential Features for Real-Time UAV Tracking

Cited by 12 publications

References 30 publications

Unsupervised Nighttime Object Tracking Based on Transformer and Domain Adaptation Fusion Network

Unsupervised Nighttime Object Tracking Based on Transformer and Domain Adaptation Fusion Network

From Augmentation to Inpainting: Improving Visual SLAM With Signal Enhancement Techniques and GAN-Based Image Inpainting

Exploring Knowledge Transfer and Distillation in Deep Learning

Contact Info

Product

Resources

About