3D Human Body Reconstruction from a Single Image via Volumetric Regression

TL;DR

This paper introduces an end-to-end CNN approach for 3D human body reconstruction from a single image, capable of handling pose variation and reconstructing occluded parts without relying on shape models.

Contribution

It presents a novel volumetric regression method that directly predicts 3D geometry from various input types, including images, landmarks, and segmentation masks.

Findings

Effective reconstruction of 3D human bodies from single images.

Handles pose variation and occlusions.

Does not require shape model fitting.

Abstract

This paper proposes the use of an end-to-end Convolutional Neural Network for direct reconstruction of the 3D geometry of humans via volumetric regression. The proposed method does not require the fitting of a shape model and can be trained to work from a variety of input types, whether it be landmarks, images or segmentation masks. Additionally, non-visible parts, either self-occluded or otherwise, are still reconstructed, which is not the case with depth map regression. We present results that show that our method can handle both pose variation and detailed reconstruction given appropriate datasets for training.