Adversarial Training for Adverse Conditions: Robust Metric Localisation Using Appearance Transfer @ICRA-cg8kk
Adversarial Training for Adverse Conditions: Robust Metric Localisation Using Appearance Transfer  @ICRA-cg8kk
Uploaded May 2018 | Updated November 2026, 2 days ago
ICRA 2018 Spotlight Video
Interactive Session Tue AM Pod R.5
Authors: Porav, Horia; Maddern, Will; Newman, Paul
Title: Adversarial Training for Adverse Conditions: Robust Metric Localisation Using Appearance Transfer

Abstract:
We present a method of improving visual place recognition and metric localisation under very strong appearance change. We learn an invertable generator that can transform the conditions of images, e.g. from day to night, summer to winter etc. This image transforming filter is explicitly designed to aid and abet feature-matching using a new loss based on SURF detector and dense descriptor maps. A network is trained to output synthetic images, features and descriptors given only an input RGB image, and these generated outputs are used to localize the robot against a previously built map using traditional sparse matching approaches. We benchmark our results using multiple traversals of the Oxford RobotCar Dataset over a year-long period, using one traversal as a map and the other to localise. We show that this method significantly improves place recognition and localisation under changing and adverse conditions, while reducing the number of mapping runs needed to successfully achieve reliable localisation.
Adversarial Training for Adverse Conditions: Robust Metric Localisation Using Appearance TransferNegotiating with a Robot: Analysis of Regulatory Focus BehaviorWorkspace, Transmissibility and Dynamics of a New 3T3R Parallel Pick-And-Place Robot with High RotatMaestROB: A Robotics Framework for Integrated Orchestration of Low-Level Control and High-Level ReasFull Body Attitude Stabilizer for Wheel-Legged Quadruped RobotsSlip Detection with Combined Tactile and Visual InformationOptical Sensing and Control Methods for Soft Pneumatically Actuated Robotic ManipulatorsApproximate Branch and Bound for Fast, Risk-Bound Stochastic Path PlanningAnalyzing and Improving Cartesian Stiffness Control Stability of Series Elastic Tendon-Driven RobotiSurface-Based Exploration for Autonomous 3D ModelingFast Object Learning and Dual-Arm Coordination for Cluttered Stowing, Picking, and PackingRealtime Planning for High-DOF Deformable Bodies Using Two-Stage Learning
ICRA 2018 |

Adversarial Training for Adverse Conditions: Robust Metric Localisation Using Appearance Transfer

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER