Pixel-Wise Prediction based Visual Odometry via Uncertainty Estimation

TL;DR

This paper presents PWVO, a dense visual odometry method that uses pixel-wise predictions and uncertainty estimation to improve translation and rotation estimation from input observations.

Contribution

It introduces a novel pixel-wise prediction framework with uncertainty-based noise filtering and a synthetic data generation workflow for training.

Findings

PWVO achieves favorable accuracy in visual odometry tasks.

Uncertainty maps effectively identify noisy regions in observations.

The method demonstrates robustness through comprehensive analysis.

Abstract

This paper introduces pixel-wise prediction based visual odometry (PWVO), which is a dense prediction task that evaluates the values of translation and rotation for every pixel in its input observations. PWVO employs uncertainty estimation to identify the noisy regions in the input observations, and adopts a selection mechanism to integrate pixel-wise predictions based on the estimated uncertainty maps to derive the final translation and rotation. In order to train PWVO in a comprehensive fashion, we further develop a data generation workflow for generating synthetic training data. The experimental results show that PWVO is able to deliver favorable results. In addition, our analyses validate the effectiveness of the designs adopted in PWVO, and demonstrate that the uncertainty maps estimated by PWVO is capable of capturing the noises in its input observations.