Look Ma, No Ground Truth! Ground-Truth-Free Tuning of Structure from Motion and Visual SLAM

TL;DR

This paper introduces a ground-truth-free evaluation method for SfM and VSLAM that uses sensitivity estimation from noisy inputs, enabling scalable, self-supervised tuning without relying on costly geometric ground truth.

Contribution

It presents a novel GTF evaluation approach that correlates with traditional benchmarks and facilitates hyperparameter tuning without ground truth data.

Findings

Strong correlation with ground-truth benchmarks

Enables hyperparameter tuning without geometric ground truth

Supports self-supervised and online tuning methods

Abstract

Evaluation is critical to both developing and tuning Structure from Motion (SfM) and Visual SLAM (VSLAM) systems, but is universally reliant on high-quality geometric ground truth -- a resource that is not only costly and time-intensive but, in many cases, entirely unobtainable. This dependency on ground truth restricts SfM and SLAM applications across diverse environments and limits scalability to real-world scenarios. In this work, we propose a novel ground-truth-free (GTF) evaluation methodology that eliminates the need for geometric ground truth, instead using sensitivity estimation via sampling from both original and noisy versions of input images. Our approach shows strong correlation with traditional ground-truth-based benchmarks and supports GTF hyperparameter tuning. Removing the need for ground truth opens up new opportunities to leverage a much larger number of dataset sources, and for self-supervised and online tuning, with the potential for a data-driven breakthrough analogous to what has occurred in generative AI.