Tsung-Han Ho, Chen-Yin Yu, Tsai-Yen Ko, and Wei-Ta Chu
Multimedia and Computer Vision Laboratory
Dept. of Computer Science and Information Engineering
National Cheng Kung University
1. Introduction
We propose a multi-scale thermal-to-visible face synthesis system to achieve thermal face recognition. A generative adversarial network is constructed by one generator that transforms a given thermal face into a face in the visible spectrum, and three discriminators that consider multi-scale feature matching and high-frequency components, respectively. In addition, we provide a new paired thermal-visible face dataset called VTF that mainly contains Asian subjects captured in various visual conditions. This new dataset not only poses technical challenges to thermal face recognition, but also enables us to point out the race bias issue in current thermal face recognition methods. Overall, the proposed system achieves the state-of-the-art performance in both the EURECOM and NCKU-VTF datasets.
2. The NCKU-VTF Dataset
This visible and thermal paired face database is with ground truths of identity and facial landmarks. The range of the participants' ages is between 19 to 57, the gender ratio of males to females is 1 to 3, and the ethnic distribution includes Indonesian, Colombian, Iranian, and Taiwanese. The collection settings mainly refer to the EURECOM dataset with some extensions.
Each subject was captured two times separated by at least 7 days but not exceeding two weeks. We say images are captured in two runs. In each run 30 pairs of visual and thermal face images were captured per person. The total number of face images in the VTF dataset is 6,000.
Dataset download:
).3. Citation
Please cite our work if you utilize this dataset.
Tsung-Han Ho, Chen-Yin Yu, Tsai-Yen Ko, and Wei-Ta Chu, "The NCKU-VTF Dataset and a Multi-scale Thermal-to-Visible Face Synthesis System," Proceedings of International Conference on Multimedia Modeling, pp. 463-475, 2023.
Last Updated: August 31, 2023