Zewei He

Real-world images plagued by both raindrops and reflections finally get a dedicated benchmark dataset (RDRF) and a diffusion-based model (DiffUR$^3$) that actually works.

Synthesizing training data with foundation models and attending to wavelet domains can dramatically boost anomaly detection, even without fine-tuning or class-specific training.

Achieve superior single image reflection removal by closing the semantic gap between pre-trained models and reflection removal models, and unifying reflection labels across synthetic and real-world data.