AI For Robotics Navigation L

Autonomous navigation is a critical ability for a robot to be able to move and operate efficiently. However, constructing the maps that robots use, such as 3D point clouds and occupancy grids, is rather complicated. They require a number of different sensor data as well as frequent updates when the robots operate in dynamic, changing environments and public spaces such as shopping malls or airports. We believe that AI can help to construct and maintain new kinds of maps or representations of the robot’s environment, as well as intelligent perception systems that enable robust navigation.

Going one step further, a robot should be able to learn how to navigate in an environment without being dependent upon precise sensor readings and maps. Ad-hoc navigation decisions based on the robot’s observation of its environment (e.g. with a camera), the given task, and past experience could replace the traditional path planning and path following approach.




Title Description Year Code/data Related papers/Blog
Large-scale localization Large-scale localization datasets in crowded indoor spaces. 2019 Dataset CVPR 2021
Virtual KITTI 2 Updated photo-realistic synthetic video dataset designed to learn and evaluate computer vision models for several video understanding tasks: object detection and multi-object tracking, scene-level and instance-level semantic segmentation, optical flow, and depth estimation. 2020 Download dataset arXiv/
Virtual Gallery Dataset Synthetic dataset targeting challenges such as varying lighting conditions and different occlusion levels for tasks such as depth estimation, instance segmentation and visual localization. 2019 Dataset CVPR 2019


Title Description Year Code/data Related papers/Blog
Kapture Kapture is a file format as well as a set of tools for manipulating datasets, and in particular Visual Localization and Structure from Motion data. 2020 GitHub arXiv/
Kapture localization A toolbox with various localization related algorithms (mapping, localization, benchmarking IR for VL). Relies strongly on the kapture format for data representation and manipulation. 2020 GitHub 3DV 2020
R2D2 Reliable and Repeatable Detector and Descriptor. 2019 Github (Code and data) NeurIPs 2019/
Deep Image Retrieval End-to-end learning of deep visual representations for image retrieval. 2017 Models and scripts of papers ICCV 2019/
IJCV 2017/


Localization Datasets in Crowded Indoor Spaces
NAVER LABS has made available five new indoor datasets for large scale visual localization in crowded public spaces. Blog article by Donghwan Lee, Soohyun Ryu, Suyong Yeon, Yonghan Lee, Deokhwa Kim, Cheolho Han, Yohann Cabon, Philippe Weinzaepfel, Nicolas Guerin, Gabriela Csurka Khedari and Martin Humenberger
One Method One Pipeline Blog Image
Using the KAPTURE pipeline and robust R2D2 method, we ranked 1st, 2nd and 4th in the VisLocOdomMap workshop challenge. Blog article by Martin Humenberger
An overview of the evolution of local features. Blog article by Gabriela Csurka Khedari, Christopher Dance and Martin Humenberger
Open source release of the structure from motion and visual localization data format kapture. Blog article by Martin Humenberger
Fast Adaptation Blog Image 1
Our navigation system enables robots to adapt to specific real-world environments and use cases with only small amounts of human preference data. Blog article by Christopher Dance and Jinyoung Choi
Combining keypoint reliability in an image as part of the keypoint detection problem significantly improves feature matching. Blog article by Jérome Revaud, Philippe Weinzaepfel, Cesar De Souza and Martin Humenberger
New release of the popular synthetic image dataset for training and testing. Blog article by Martin Humenberger, Yohann Cabon and Naila Murray
Visual Localization
The first deep regression-based method that generalizes to larger environments, performs better than state-of-the-art with little training data and is robust to occlusion. Virtual Gallery dataset provided. Blog article by Philippe Weinzaepfel, Gabriela Csurka Khedari, Yohann Cabon and Martin Humenberger
A system that correctly detects when places have changed to automatically update complex indoor maps. Dataset available for research. Blog article by Jérome Revaud

Related Content

This web site uses cookies for the site search, to display videos and for aggregate site analytics.

Learn more about these cookies in our privacy notice.


Cookie settings

You may choose which kind of cookies you allow when visiting this website. Click on "Save cookie settings" to apply your choice.

FunctionalThis website uses functional cookies which are required for the search function to work and to apply for jobs and internships.

AnalyticalOur website uses analytical cookies to make it possible to analyse our website and optimize its usability.

Social mediaOur website places social media cookies to show YouTube and Vimeo videos. Cookies placed by these sites may track your personal data.