Hyperparameter-Free Losses for Model-Based Monocular Reconstruction

TL;DR

This paper introduces hyperparameter-free loss functions for monocular 3D reconstruction that simplify optimization, reduce complexity, and improve alignment accuracy without extra annotations, benefiting applications like augmented reality.

Contribution

It presents novel hyperparameter-free losses leveraging geometry, enabling joint optimization of shape and pose, and introduces an implicit regularization technique based on virtual projections.

Findings

Reduced optimization time and complexity.

Effective for precise geometry-image alignment.

Validated on large-scale datasets with 3D ground truth.

Abstract

This work proposes novel hyperparameter-free losses for single view 3D reconstruction with morphable models (3DMM). We dispense with the hyperparameters used in other works by exploiting geometry, so that the shape of the object and the camera pose are jointly optimized in a sole term expression. This simplification reduces the optimization time and its complexity. Moreover, we propose a novel implicit regularization technique based on random virtual projections that does not require additional 2D or 3D annotations. Our experiments suggest that minimizing a shape reprojection error together with the proposed implicit regularization is especially suitable for applications that require precise alignment between geometry and image spaces, such as augmented reality. We evaluate our losses on a large scale dataset with 3D ground truth and publish our implementations to facilitate reproducibility and public benchmarking in this field.