Unsupervised Cross-Modal Alignment for Multi-Person 3D Pose Estimation

TL;DR

This paper introduces a fast, deployment-friendly bottom-up framework for multi-person 3D pose estimation that learns a shared latent space through cross-modal alignment, eliminating the need for paired supervision and keypoint grouping.

Contribution

It proposes a novel neural representation for multi-person 3D poses, a cross-modal training paradigm without paired annotations, and a generative pose embedding that improves speed and accuracy.

Findings

Achieves state-of-the-art results among bottom-up methods.

Generalizes well to in-the-wild images.

Offers a superior speed-performance trade-off.

Abstract

We present a deployment friendly, fast bottom-up framework for multi-person 3D human pose estimation. We adopt a novel neural representation of multi-person 3D pose which unifies the position of person instances with their corresponding 3D pose representation. This is realized by learning a generative pose embedding which not only ensures plausible 3D pose predictions, but also eliminates the usual keypoint grouping operation as employed in prior bottom-up approaches. Further, we propose a practical deployment paradigm where paired 2D or 3D pose annotations are unavailable. In the absence of any paired supervision, we leverage a frozen network, as a teacher model, which is trained on an auxiliary task of multi-person 2D pose estimation. We cast the learning as a cross-modal alignment problem and propose training objectives to realize a shared latent space between two diverse modalities. We aim to enhance the model's ability to perform beyond the limiting teacher network by enriching the latent-to-3D pose mapping using artificially synthesized multi-person 3D scene samples. Our approach not only generalizes to in-the-wild images, but also yields a superior trade-off between speed and performance, compared to prior top-down approaches. Our approach also yields state-of-the-art multi-person 3D pose estimation performance among the bottom-up approaches under consistent supervision levels.