AutoLumNet: A Bi-Branch Exposure-Aware Network for Low- and High-Exposure Image Enhancement
Published in International Conference on Learning Representations (ICLR), 2026 – Under Review, 2026
AutoLumNet proposes a unified, bi-branch architecture for enhancing both low-exposure and high-exposure images. One branch focuses on boosting underexposed regions, while the other suppresses overexposed areas, with exposure-aware fusion combining their outputs.
The method is evaluated on custom and public datasets, demonstrating competitive or superior performance to state-of-the-art low-light and exposure-correction models, while maintaining structural consistency and natural color appearance.
Recommended citation: Airin Akter Tania, Md Raihan Khan, and Mohiuddin Ahmad, "AutoLumNet: A Bi-Branch Exposure-Aware Network for Low- and High-Exposure Image Enhancement," ICLR 2026 (under review).
Download Paper