The study focuses on increasing the resolution of depth maps obtained from 3D cameras using structured light technology. Two deep learning models, FDSR and DKN, are modified to work with high-resolution data. The models are trained on a custom dataset of 1200 3D scans. The approach for depth map upsampling reduces the processing time by first downsampling a high-resolution depth map, performing processing steps at the lower resolution, and then upsampling the resulting depth map. The FDSR model is faster while the DKN model provides results with higher precision.
Publication date: 14 Nov 2023
Project Page: Not provided
Paper: https://arxiv.org/pdf/2311.07432
