3D-LMNet: Latent Embedding Matching for Accurate and Diverse 3D Point Cloud Reconstruction from a Single Image

TL;DR

This paper introduces 3D-LMNet, a novel method for single-image 3D point cloud reconstruction that leverages latent embedding matching and probabilistic modeling to produce accurate, diverse, and plausible 3D reconstructions.

Contribution

The paper presents a two-stage approach combining a 3D auto-encoder with a learned mapping from images to latent space, enabling multiple plausible reconstructions with diversity loss.

Findings

Outperforms state-of-the-art on real and synthetic datasets

Generates multiple diverse reconstructions for a single input image

Effectively models uncertainty in 3D reconstruction

Abstract

3D reconstruction from single view images is an ill-posed problem. Inferring the hidden regions from self-occluded images is both challenging and ambiguous. We propose a two-pronged approach to address these issues. To better incorporate the data prior and generate meaningful reconstructions, we propose 3D-LMNet, a latent embedding matching approach for 3D reconstruction. We first train a 3D point cloud auto-encoder and then learn a mapping from the 2D image to the corresponding learnt embedding. To tackle the issue of uncertainty in the reconstruction, we predict multiple reconstructions that are consistent with the input view. This is achieved by learning a probablistic latent space with a novel view-specific diversity loss. Thorough quantitative and qualitative analysis is performed to highlight the significance of the proposed approach. We outperform state-of-the-art approaches on the task of single-view 3D reconstruction on both real and synthetic datasets while generating multiple plausible reconstructions, demonstrating the generalizability and utility of our approach.