Deep Non-rigid Structure-from-Motion Revisited: Canonicalization and Sequence Modeling

TL;DR

This paper improves deep non-rigid structure-from-motion by introducing a per-sequence canonicalization method and a sequence modeling approach that leverages temporal information and subspace constraints, leading to better 3D reconstructions.

Contribution

It presents a simple per-sequence canonicalization technique and a sequence modeling framework that enhances deep NRSfM performance over existing methods.

Findings

Achieved more accurate 3D reconstructions on standard datasets.

Demonstrated the effectiveness of sequence modeling with temporal and subspace constraints.

Outperformed previous deep NRSfM approaches in experiments.

Abstract

Non-Rigid Structure-from-Motion (NRSfM) is a classic 3D vision problem, where a 2D sequence is taken as input to estimate the corresponding 3D sequence. Recently, the deep neural networks have greatly advanced the task of NRSfM. However, existing deep NRSfM methods still have limitations in handling the inherent sequence property and motion ambiguity associated with the NRSfM problem. In this paper, we revisit deep NRSfM from two perspectives to address the limitations of current deep NRSfM methods : (1) canonicalization and (2) sequence modeling. We propose an easy-to-implement per-sequence canonicalization method as opposed to the previous per-dataset canonicalization approaches. With this in mind, we propose a sequence modeling method that combines temporal information and subspace constraint. As a result, we have achieved a more optimal NRSfM reconstruction pipeline compared to previous efforts. The effectiveness of our method is verified by testing the sequence-to-sequence deep NRSfM pipeline with corresponding regularization modules on several commonly used datasets.