logic112358 / rt-can

To facilitate the research of invisible gas detection, we introduce Gas-DB, an extensive open-source gas detection database including about 1.3K well-annotated RGB-thermal images with eight variant collection scenes.

Invisible Gas Detection: An RGB-Thermal Cross Attention Network and A New Benchmark

Paper | Gas-DB

Authors

• Jue Wang*^1,3, Yuxiang Lin*², Qi Zhao³, Dong Luo³, Shuaibao Chen³, Wei Chen³, Xiaojiang Peng²
• ¹Southern University of Science and Technology, ²Shenzhen Technology University, ³Shenzhen Institute of Technology, CAS [Jue and Yuxiang contribute equally to this work.]

If you are interested in our work, please star ⭐ our project.

Training dataset preparation

• Prepare our Gas-DB dataset: please download in Gas-DB.

Code will be made available soon. Stay tuned!

RGB-Thermal Cross Attention Network

Illustration the architecture of RGB-Thermal Two Stream Cross Attention Network. (a) Two stream RGB-ThermaR Cl Encoder, (b) Cascaded Decoder.

An overview of our Gas-DB  

This figure shows an overview of our Gas-DB, containing 8 kinds of scenery, containing sunny, rainy, double leakage, nearly leakage, further leakage, overlook, simple background, and complex background. The last one is the original gas image without manually annotating.

The comparision of the GasVid and our Gas-DB.

The visualization of the prediction comparisons from different methods

The visualization of the prediction comparisons from different methods, according to the rows from top to bottom in order: RGB; Thermal; Ground Truth; PSPNet; Segformer; YOLOv5; MFNet; EAEFNet; Ours.

Contact   

For any question, feel free to email [email protected] and [email protected].

Citation

@article{RT-CAN,
  title={Invisible Gas Detection: An RGB-Thermal Cross Attention Network and A New Benchmark},
  author={Wang, Jue and Lin, Yuxiang and Zhao, Qi and Luo, Dong and Chen, Shuaibao and Chen, Wei and Peng, Xiaojiang},
  journal={arXiv preprint arXiv:2403.17712},
  year={2024}
}