To handle the issue, we propose a new framework, coined robust model adaptation (RoMA), based on gradient-based optimization of inputs over the DNN. Specifically, it consists oftwosteps: (a)apre-training strategytorobustly train theproxy model and (b) a novel adaptation procedure of the proxy model to have robust estimates for a specific set of candidate solutions. At ahigh level, our scheme utilizes thelocal smoothness priorto overcome the brittleness of the DNN.

To handle the issue, we propose a new framework, coined robust model adaptation (RoMA), based on gradient-based optimization of inputs over the DNN. Specifically, it consists oftwosteps: (a)apre-training strategytorobustly train theproxy model and (b) a novel adaptation procedure of the proxy model to have robust estimates for aspecific set ofcandidate solutions. Atahigh level, our scheme utilizes thelocal smoothness priorto overcome the brittleness of the DNN.

Image restoration (IR) stands as a fundamental task within low-level computer vision, aiming to enhance the quality of images affected by numerous factors, including noise, blur,lowresolution, compression artifacts, mosaic patterns, adverse weather conditions, and other forms of distortion. This capability holds broad utility across various domains, facilitating information recovery in medical imaging, surveillance, and satellite imagery. Furthermore, it bolsters downstream vision tasks like object detection, recognition, and tracking [74, 60].