On Reproducing Semi-dense Depth Map Reconstruction using Deep Convolutional Neural Networks with Perceptual Loss

P. 1080–1084.

Makarov I., Dmitrii Maslov, Gerasimova O., Vladimir Aliev, Alisa Korinevskaya, Sharma U., Wang H.

In our recent papers, we proposed a new family of residual convolutional neural networks trained for semi-dense and sparse depth reconstruction without use of RGB channel. The proposed models can be used in low-resolution depth sensors or SLAM methods estimating partial depth with certain distributions. We proposed using perceptual loss for training depth reconstruction in order to better preserve edge structure and reduce over-smoothness of models trained on MSE loss alone. 

This paper contains reproducibility companion guide on training, running and evaluating suggested methods, while also presenting links on further studies in view of reviewers comments and related problems of depth reconstruction.

Keywords: computer vision, Augmented Reality, convolutional neural networks, Autonomous vehicles, Depth map, Mixed Reality

Publication based on the results of: Discovering and Representing Knowledge for Recommender Systems (2019)

In book

Proceedings of 27th ACM International Conference on Multimedia

NY: ACM, 2019.