Hanyang University Team's AI Research on Low-Light Object Detection Accepted by Top Conference ICLR 2026

Hanyang University's researchers are making significant strides in the field of artificial intelligence, with a recent breakthrough in low-light object detection earning them a spot at the prestigious ICLR 2026 conference. The team, led by Professor Hwang Soon-min from the Department of Future Automotive Engineering, has developed an innovative 'Self-guided low-light object detection framework.' What makes this achievement particularly noteworthy is the prominent role played by undergraduate and master's students, who were instrumental in developing the technology. This framework tackles a persistent challenge in AI: the degradation of object detection performance in dimly lit or noisy environments. Unlike previous methods that often amplify noise or increase computational load, the Hanyang team's approach utilizes a unique 'Fourier-based image fusion' technique. This method intelligently separates and recombines image components in the frequency domain, preserving essential structural information while effectively suppressing noise. The result is a dramatic improvement in accuracy, demonstrated by a significant leap in performance on benchmark datasets. Crucially, this enhanced capability comes without any added computational overhead during the inference stage, making it highly practical for real-time applications such as autonomous vehicles and advanced surveillance systems. This success highlights the cutting-edge research being conducted at Hanyang University and its potential to drive advancements across various industries.

“

The researchers who led the successful publication of the ICLR paper after leading the CVPR 2025 Workshop 'Waymo Vision-based End-to-End Driving' challenge to 3rd place have proven the lab's unique technological prowess.