Leveraging the Learnable Vertex-Vertex Relationship to Generalize Human Pose and Mesh Reconstruction for In-the-Wild Scenes

TL;DR

MeshLeTemp introduces a learnable human mesh template that captures vertex interactions and adapts to diverse images, significantly improving 3D human pose and mesh reconstruction in wild scenes.

Contribution

The paper proposes a learnable template mesh that encodes vertex relationships and adapts to different poses and shapes, advancing prior fixed-template methods.

Findings

Demonstrates improved generalization to unseen scenarios.

Outperforms previous methods in wild scene reconstructions.

Validates effectiveness through extensive experiments.

Abstract

We present MeshLeTemp, a powerful method for 3D human pose and mesh reconstruction from a single image. In terms of human body priors encoding, we propose using a learnable template human mesh instead of a constant template as utilized by previous state-of-the-art methods. The proposed learnable template reflects not only vertex-vertex interactions but also the human pose and body shape, being able to adapt to diverse images. We conduct extensive experiments to show the generalizability of our method on unseen scenarios.