[1607.07988] ATGV-Net: Accurate Depth Super-Resolutionopen searchopen navigation menucontact arXivsubscribe to arXiv mailings

In this work we present a novel approach for single depth map super-resolution. Modern consumer depth sensors, especially Time-of-Flight sensors, produce dense depth measurements, but are affected by noise and have a low lateral resolution. We propose a method that combines the benefits of recent advances in machine learning based single image super-resolution, i.e. deep convolutional networks, with a variational method to recover accurate high-resolution depth maps. In particular, we integrate a variational method that models the piecewise affine structures apparent in depth data via an anisotropic total generalized variation regularization term on top of a deep network. We call our method ATGV-Net and train it end-to-end by unrolling the optimization procedure of the variational method. To train deep networks, a large corpus of training data with accurate ground-truth is required. We demonstrate that it is feasible to train our method solely on synthetic data that we generate in larg

1 mentions: @ankurhandos
Date: 2020/11/21 08:21

Referring Tweets

@ankurhandos RAFT iteratively refines optical flow with GRU. Differentiable primal-dual is another interesting approach to iteratively refine solutions t.co/xmxSpKuH36 t.co/iTNRzfyiHU

Related Entries

Read more Learning the signatures of the human grasp using a scalable tactile glove
0 users, 1 mentions 2020/03/15 17:21
Read more [1907.04502] DeepXDE: A deep learning library for solving differential equationscontact arXivarXiv T...
0 users, 1 mentions 2020/04/01 17:21
Read more Path-Space Differentiable Rendering
0 users, 1 mentions 2020/05/15 17:21
Read more Portrait Shadow Manipulation
0 users, 1 mentions 2020/05/15 20:21
Read more [1811.12463] Fast and Flexible Indoor Scene Synthesis via Deep Convolutional Generative Modelsopen s...
0 users, 1 mentions 2020/09/06 15:51