Multi-task Learning with 3D-Aware Regularization

TL;DR

This paper introduces a 3D-aware regularizer for multi-task learning in computer vision, improving task correlation modeling by projecting features into a shared 3D space and enhancing performance across multiple benchmarks.

Contribution

The paper proposes a novel 3D-aware regularizer that interfaces multiple tasks via a shared 3D feature space, improving multi-task learning performance.

Findings

Improves multi-task learning performance on NYUv2 and PASCAL-Context.

Architecture-agnostic method adaptable to various backbones.

Enhances correlation modeling between tasks through 3D regularization.

Abstract

Deep neural networks have become a standard building block for designing models that can perform multiple dense computer vision tasks such as depth estimation and semantic segmentation thanks to their ability to capture complex correlations in high dimensional feature space across tasks. However, the cross-task correlations that are learned in the unstructured feature space can be extremely noisy and susceptible to overfitting, consequently hurting performance. We propose to address this problem by introducing a structured 3D-aware regularizer which interfaces multiple tasks through the projection of features extracted from an image encoder to a shared 3D feature space and decodes them into their task output space through differentiable rendering. We show that the proposed method is architecture agnostic and can be plugged into various prior multi-task backbones to improve their performance; as we evidence using standard benchmarks NYUv2 and PASCAL-Context.