LLVIP: A Visible-infrared Paired Dataset for Low-light Vision

Jia, Xinyu; Zhu, Chuang; Li, Minzhen; Tang, Wenqi; Zhou, Wenli

doi:10.48550/arxiv.2108.10831

Cited by 3 publications

(2 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The combined information improves the robot's field of view in the environment. This integration strives to capitalize on the strengths of both sensor types [32,35], overcoming the limitations posed by LiDAR's temporal resolution while leveraging its wide field of view for robust obstacle detection and avoidance.…”

Section: Sensor Fusionmentioning

confidence: 99%

“…This approach includes capturing a wide range of obstacles found in MRO hangars, including tools, vehicles, and structural elements in the real world and in simulation, and using these datasets to train the model. LLVIP [35] public datasets were also integrated to expand the capability of recognizing objects in diverse scenarios considering low-light conditions, which are common in certain areas of the MRO hangar or during specific times of the day. For data fusion, the DenseFuse network was first used to fuse RGB and thermal images from real-world and LLVIP datasets to expand the model's capability to accurately detect obstacles in low-light conditions.…”

Section: Obstacle Detection Modulementioning

confidence: 99%

See 1 more Smart Citation

Improved Hybrid Model for Obstacle Detection and Avoidance in Robot Operating System Framework (Rapidly Exploring Random Tree and Dynamic Windows Approach)

Adiuku,

Avdelidis,

Tang

et al. 2024

Sensors

View full text Add to dashboard Cite

The integration of machine learning and robotics brings promising potential to tackle the application challenges of mobile robot navigation in industries. The real-world environment is highly dynamic and unpredictable, with increasing necessities for efficiency and safety. This demands a multi-faceted approach that combines advanced sensing, robust obstacle detection, and avoidance mechanisms for an effective robot navigation experience. While hybrid methods with default robot operating system (ROS) navigation stack have demonstrated significant results, their performance in real time and highly dynamic environments remains a challenge. These environments are characterized by continuously changing conditions, which can impact the precision of obstacle detection systems and efficient avoidance control decision-making processes. In response to these challenges, this paper presents a novel solution that combines a rapidly exploring random tree (RRT)-integrated ROS navigation stack and a pre-trained YOLOv7 object detection model to enhance the capability of the developed work on the NAV-YOLO system. The proposed approach leveraged the high accuracy of YOLOv7 obstacle detection and the efficient path-planning capabilities of RRT and dynamic windows approach (DWA) to improve the navigation performance of mobile robots in real-world complex and dynamically changing settings. Extensive simulation and real-world robot platform experiments were conducted to evaluate the efficiency of the proposed solution. The result demonstrated a high-level obstacle avoidance capability, ensuring the safety and efficiency of mobile robot navigation operations in aviation environments.

show abstract

Section: Sensor Fusionmentioning

confidence: 99%

Section: Obstacle Detection Modulementioning

confidence: 99%

Improved Hybrid Model for Obstacle Detection and Avoidance in Robot Operating System Framework (Rapidly Exploring Random Tree and Dynamic Windows Approach)

Adiuku,

Avdelidis,

Tang

et al. 2024

Sensors

View full text Add to dashboard Cite

show abstract

Flexible thermal camera solution for Smart city people detection and counting

Collini,

Palesi,

Nesi

et al. 2023

Multimed Tools Appl

View full text Add to dashboard Cite

Tourism management plays an important role in the context of Smart Cities. In this work, we have used thermal cameras for the development of an Object Detection solution in pedestrian areas. The solution can classify people, bikes, strollers, and count people in Real-Time by using telephoto and wide-angle thermal cameras, in hot squares where there is a relevant number of people passing by. This work has improved FASTER-R-CNN and YOLOv5 architectures with new data sets and fine-tuning approaches to enhance mean average precision and flexibility whether compared to state of the art solutions. Both top-down and bottom-up training adaptation approaches have been assessed in order to demonstrate that the proposed bottom-up approach can provide better results. Results have overcome the state-of-the-art in terms of mean Average Precision in counting (i) for relevant number of people in the scene (removing the limitation of previous state-of-the-art solutions that were set to provide good precision up to 10 people) and (ii) in terms of flexibility with respect to different kinds of camera and resolutions. The resulting model can produce results also when executed on thermal camera and in Real-Time on industrial PC of mid-level. The proposed solution has been developed and validated in the framework of the Herit-Data EC project and it has exploited the Snap4City platform for the final collection of data results, monitoring and their publication on real time dashboards.

show abstract