
Visible-thermal pedestrian detection via unsupervised transfer learning
- Author
- Chengjin Lyu (UGent) , Patrick Heyer Wollenberg (UGent) , Asad Munir, Ljiljana Platisa (UGent) , Christian Micheloni, Bart Goossens (UGent) and Wilfried Philips (UGent)
- Organization
- Abstract
- Recently, pedestrian detection using visible-thermal pairs plays a key role in around-the-clock applications, such as public surveillance and autonomous driving. However, the performance of a well-trained pedestrian detector may drop significantly when it is applied to a new scenario. Normally, to achieve a good performance on the new scenario, manual annotation of the dataset is necessary, while it is costly and unscalable. In this work, an unsupervised transfer learning framework is proposed for visible-thermal pedestrian detection tasks. Given well-trained detectors from a source dataset, the proposed framework utilizes an iterative process to generate and fuse training labels automatically, with the help of two auxiliary single-modality detectors (visible and thermal). To achieve label fusion, the knowledge of daytime and nighttime is adopted to assign priorities to labels according to their illumination, which improves the quality of generated training labels. After each iteration, the existing detectors are updated using new training labels. Experimental results demonstrate that the proposed method obtains state-of-the-art performance without any manual training labels on the target dataset.
- Keywords
- Deep neural networks, Domain adaption, Unsupervised transfer learning, Pedestrian detection
Downloads
-
(...).pdf
- full text (Published version)
- |
- UGent only
- |
- |
- 797.88 KB
-
BAI21-521.pdf
- full text (Accepted manuscript)
- |
- open access
- |
- |
- 737.92 KB
Citation
Please use this url to cite or link to this publication: http://hdl.handle.net/1854/LU-8689648
- MLA
- Lyu, Chengjin, et al. “Visible-Thermal Pedestrian Detection via Unsupervised Transfer Learning.” 2021 5TH INTERNATIONAL CONFERENCE ON INNOVATION IN ARTIFICIAL INTELLIGENCE (ICIAI 2021), Association for Computing Machinery (ACM), 2021, pp. 158–63, doi:10.1145/3461353.3461369.
- APA
- Lyu, C., Heyer Wollenberg, P., Munir, A., Platisa, L., Micheloni, C., Goossens, B., & Philips, W. (2021). Visible-thermal pedestrian detection via unsupervised transfer learning. 2021 5TH INTERNATIONAL CONFERENCE ON INNOVATION IN ARTIFICIAL INTELLIGENCE (ICIAI 2021), 158–163. https://doi.org/10.1145/3461353.3461369
- Chicago author-date
- Lyu, Chengjin, Patrick Heyer Wollenberg, Asad Munir, Ljiljana Platisa, Christian Micheloni, Bart Goossens, and Wilfried Philips. 2021. “Visible-Thermal Pedestrian Detection via Unsupervised Transfer Learning.” In 2021 5TH INTERNATIONAL CONFERENCE ON INNOVATION IN ARTIFICIAL INTELLIGENCE (ICIAI 2021), 158–63. Association for Computing Machinery (ACM). https://doi.org/10.1145/3461353.3461369.
- Chicago author-date (all authors)
- Lyu, Chengjin, Patrick Heyer Wollenberg, Asad Munir, Ljiljana Platisa, Christian Micheloni, Bart Goossens, and Wilfried Philips. 2021. “Visible-Thermal Pedestrian Detection via Unsupervised Transfer Learning.” In 2021 5TH INTERNATIONAL CONFERENCE ON INNOVATION IN ARTIFICIAL INTELLIGENCE (ICIAI 2021), 158–163. Association for Computing Machinery (ACM). doi:10.1145/3461353.3461369.
- Vancouver
- 1.Lyu C, Heyer Wollenberg P, Munir A, Platisa L, Micheloni C, Goossens B, et al. Visible-thermal pedestrian detection via unsupervised transfer learning. In: 2021 5TH INTERNATIONAL CONFERENCE ON INNOVATION IN ARTIFICIAL INTELLIGENCE (ICIAI 2021). Association for Computing Machinery (ACM); 2021. p. 158–63.
- IEEE
- [1]C. Lyu et al., “Visible-thermal pedestrian detection via unsupervised transfer learning,” in 2021 5TH INTERNATIONAL CONFERENCE ON INNOVATION IN ARTIFICIAL INTELLIGENCE (ICIAI 2021), Xiamen, China (Online), 2021, pp. 158–163.
@inproceedings{8689648, abstract = {{Recently, pedestrian detection using visible-thermal pairs plays a key role in around-the-clock applications, such as public surveillance and autonomous driving. However, the performance of a well-trained pedestrian detector may drop significantly when it is applied to a new scenario. Normally, to achieve a good performance on the new scenario, manual annotation of the dataset is necessary, while it is costly and unscalable. In this work, an unsupervised transfer learning framework is proposed for visible-thermal pedestrian detection tasks. Given well-trained detectors from a source dataset, the proposed framework utilizes an iterative process to generate and fuse training labels automatically, with the help of two auxiliary single-modality detectors (visible and thermal). To achieve label fusion, the knowledge of daytime and nighttime is adopted to assign priorities to labels according to their illumination, which improves the quality of generated training labels. After each iteration, the existing detectors are updated using new training labels. Experimental results demonstrate that the proposed method obtains state-of-the-art performance without any manual training labels on the target dataset.}}, author = {{Lyu, Chengjin and Heyer Wollenberg, Patrick and Munir, Asad and Platisa, Ljiljana and Micheloni, Christian and Goossens, Bart and Philips, Wilfried}}, booktitle = {{2021 5TH INTERNATIONAL CONFERENCE ON INNOVATION IN ARTIFICIAL INTELLIGENCE (ICIAI 2021)}}, isbn = {{9781450388634}}, keywords = {{Deep neural networks,Domain adaption,Unsupervised transfer learning,Pedestrian detection}}, language = {{eng}}, location = {{Xiamen, China (Online)}}, pages = {{158--163}}, publisher = {{Association for Computing Machinery (ACM)}}, title = {{Visible-thermal pedestrian detection via unsupervised transfer learning}}, url = {{http://doi.org/10.1145/3461353.3461369}}, year = {{2021}}, }
- Altmetric
- View in Altmetric
- Web of Science
- Times cited: