Related Experiment Video
Updated: May 15, 2025
10:14
Author Spotlight: Enhancing Neurorehabilitation Through EEG, Motor Imagery, and Virtual Reality
Published on: May 10, 2024
805
Tri-band vehicle and vessel dataset for artificial intelligence research
Yingjian Liu1, Gangnian Zhao1, Shuzhen Fan2,3
1School of Information Science and Engineering (ISE), Shandong University, Qingdao, 266237, China.
Scientific Data
|April 9, 2025
Summary
This study introduces a novel tri-band dataset for vehicle and vessel detection, crucial for advancing deep learning in autonomous systems. The dataset demonstrates strong performance with object detection algorithms, enabling better real-world applications.
Area of Science:
- Computer Vision
- Artificial Intelligence
- Remote Sensing
Background:
- Deep learning models require extensive datasets for training.
- Existing datasets may lack multi-band capabilities for comprehensive object detection.
- Autonomous driving and vessel detection benefit from advanced AI techniques.
Purpose of the Study:
- To introduce the first publicly available tri-band (visible, short-wave infrared, long-wave infrared) vehicle and vessel dataset.
- To facilitate object detection and multi-band image fusion research.
- To provide a benchmark for evaluating deep learning models in diverse spectral bands.
Main Methods:
- Collected and curated a dataset of thousands of tri-band images.
- Ensured time synchronization and field-of-view consistency across spectral bands.
- Manually labeled approximately 60% of the dataset for object instances.
- Trained and evaluated object detection algorithms like YOLOv8 and SSD.
- Performed preliminary validation of wavelet-based multi-band image fusion.
Main Results:
- Object detection models (YOLOv8, SSD) achieved mean Average Precision (mAP) above 0.6 at an IoU threshold of 0.5.
- The dataset supports robust recognition performance for vehicles and vessels.
- Demonstrated the feasibility of wavelet-based multi-band image fusion.
Conclusions:
- The presented tri-band dataset is a valuable resource for AI research in object detection and image fusion.
- The dataset's quality and features support the development of more accurate autonomous systems.
- This work represents a significant contribution to publicly available multi-band optical datasets.
