publications

2026

1. In prep.

   DOLENet: Detection-Oriented Low-Light Enhancement

   Yujie Chen and Minpo Jung

   Manuscript in preparation, 2026

   Abs Bib

   Low-light enhancement is usually optimised for human perception and then attached to a detector that had no say in the objective. This work studies enhancement whose only consumer is the detector, evaluated with multi-seed YOLOv8 experiments on ExDark, gradient-conflict analysis, and TP/FP/FN error decomposition.

   @article{chen2026dolenet,
     title = {DOLENet: Detection-Oriented Low-Light Enhancement},
     author = {Chen, Yujie and Jung, Minpo},
     journal = {Manuscript in preparation},
     year = {2026},
   }

2. MSc

   

   A Study on Deep Learning-Based Low-Light Image Enhancement

   Yujie Chen

   Youngsan University, 2026

   Abs Bib Code

   Zero-reference training, local illumination balancing and shadow recovery without additional inference cost. Introduces two exposure-aware losses on the Zero-DCE framework — a region-specific dark-region exposure loss and a highlight-preservation loss — benchmarked against Retinexformer and RUAS on both no-reference (NIQE, BRISQUE, LOE) and full-reference (PSNR, SSIM, LPIPS) metrics.

   @mastersthesis{chen2026thesis,
     title = {A Study on Deep Learning-Based Low-Light Image Enhancement},
     author = {Chen, Yujie},
     school = {Youngsan University},
     year = {2026},
   }