Modeling Global Illumination in Day-Night Image Translation Using Swin-CBAM CycleGAN

DOI

10.2991/978-94-6239-693-7_84How to use a DOI?

Keywords

Unpaired image translation; Day–night image translation; CycleGAN; Swin Transformer; Global illumination

Abstract

Global illumination variations make unpaired day–night image translation challenging. Under such conditions, convolution-based CycleGAN methods primarily focus on local texture translation, often neglecting global lighting consistency, which results in inconsistent sky–ground illumination and noisy outputs. To address this limitation, we propose an illumination aware CycleGAN that integrates global transformer-based attention with local convolutional attention mechanisms. The generator incorporates Swin Transformer blocks at the bottleneck to model long-range contextual dependencies, while Convolutional Block Attention Modules in the encoder–decoder enhance structural representation and suppress noise. A two stage training strategy is adopted, consisting of illumination focused pretraining on separate day and night domains, followed by fine-tuning on diverse real-world conditions. Experimental evaluations using FID, SSIM, PSNR, and LPIPS demonstrate consistent improvements over purely convolution based CycleGAN baselines, particularly in global lighting coherence and cycle consistency. These results indicate that incorporating global context modeling is crucial for realistic bidirectional day–night image translation.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - R. M. Nambiraj
AU  - V. Nithish Kumar
AU  - V. Subapriya
AU  - R. Sathya Bama Krishna
PY  - 2026
DA  - 2026/06/16
TI  - Modeling Global Illumination in Day-Night Image Translation Using Swin-CBAM CycleGAN
BT  - Proceedings of the International Conference on Intelligent Systems for a Sustainable Future (ISSF 2026)
PB  - Atlantis Press
SP  - 857
EP  - 868
SN  - 2589-4919
UR  - https://doi.org/10.2991/978-94-6239-693-7_84
DO  - 10.2991/978-94-6239-693-7_84
ID  - Nambiraj2026
ER  -