Self-supervised depth super-resolution with contrastive multiview pre-training

Xin Qiao, Chenyang Ge, Chaoqiang Zhao, Fabio Tosi, Matteo Poggi, Stefano Mattoccia

Abstract: Highlights•We propose a self-supervised method for depth upsampling with contrastive multiview pre-training.•In the pre-training, we develop an autoencoder-based framework with contrastive and reconstruction losses.•We propose a novel mutual modulation scheme.•Experimental results demonstrate the superiority of the proposed method against SOTA methods.