Given two input RGB-D frames of a non-rigidly moving object, we employ a convolutional neural network to predict dense correspondences and their confidences.

Jun 23, 2020 · We introduce a novel, end-to-end learnable, differentiable non-rigid tracker that enables state-of-the-art non-rigid reconstruction.

This repository contains the code for the NeurIPS 2020 paper Neural Non-Rigid Tracking, where we introduce a novel, end-to-end learnable, differentiable non- ...

Using our neural tracker in a non-rigid reconstruction framework results in 85× faster correspondence prediction and improved reconstruction performance ...

Existing non-rigid reconstruction approaches build the deformation graph incrementally, i.e., frame-by-frame, which can lead to unstable graph configurations in ...

Dec 6, 2020 · We introduce a novel, end-to-end learnable, differentiable non-rigid tracker that enables state-of-the-art non-rigid reconstruction by a ...

We introduce Neural Deformation Graphs for globally-consistent deformation tracking and 3D reconstruction of non-rigid objects.

A novel, end-to-end learnable, differentiable non-rigid tracker that enables state-of-the-art non- Rigid reconstruction and improves reconstruction ...

Neural Non-Rigid Tracking. 1. We thank all the reviewers for their feedback, and are happy that our work was found to be "quite novel" (R3), "well. 2 written ...

We propose Neural Non-Rigid Tracking, a differentiable non-rigid tracking approach that allows learning the correspondence prediction and weighting of ...