Continual Learning for Image-Based Camera Localization

08/20/2021
by   Shuzhe Wang, et al.
0

For several emerging technologies such as augmented reality, autonomous driving and robotics, visual localization is a critical component. Directly regressing camera pose/3D scene coordinates from the input image using deep neural networks has shown great potential. However, such methods assume a stationary data distribution with all scenes simultaneously available during training. In this paper, we approach the problem of visual localization in a continual learning setup – whereby the model is trained on scenes in an incremental manner. Our results show that similar to the classification domain, non-stationary data induces catastrophic forgetting in deep networks for visual localization. To address this issue, a strong baseline based on storing and replaying images from a fixed buffer is proposed. Furthermore, we propose a new sampling method based on coverage score (Buff-CS) that adapts the existing sampling strategies in the buffering process to the problem of visual localization. Results demonstrate consistent improvements over standard buffering methods on two challenging datasets – 7Scenes, 12Scenes, and also 19Scenes by combining the former scenes.

READ FULL TEXT
research
05/17/2021

Continual Learning with Echo State Networks

Continual Learning (CL) refers to a learning setup where data is non sta...
research
11/28/2017

Learning Less is More - 6D Camera Localization via 3D Surface Regression

Popular research areas like autonomous driving and augmented reality hav...
research
04/12/2022

LifeLonger: A Benchmark for Continual Disease Classification

Deep learning models have shown a great effectiveness in recognition of ...
research
10/12/2016

Image Based Camera Localization: an Overview

Recently, virtual reality, augmented reality, robotics, self-driving car...
research
08/03/2023

Efficient Model Adaptation for Continual Learning at the Edge

Most machine learning (ML) systems assume stationary and matching data d...
research
07/11/2022

Learning an evolved mixture model for task-free continual learning

Recently, continual learning (CL) has gained significant interest becaus...
research
12/06/2021

CSG0: Continual Urban Scene Generation with Zero Forgetting

With the rapid advances in generative adversarial networks (GANs), the v...

Please sign up or login with your details

Forgot password? Click here to reset