Carl Toft, Electrical Engineering

​Title: Towards Robust Visual Localization in Challenging Conditions

E-mail in advance to get the password

The PhD defence can be accessed through Zoom, and the it will open shortly before 15:00. We would kindly ask you to keep the video off and mute the microphone during the seminar. At the end of the session there will be an opportunity to ask questions through Zoom. In case there will be any updates about the event, these will be posted on this website.
Carl Toft is a PhD student in the research group Computer vision and medical image analysis
Opponent is Dr. Vladlen Koltun, Chief Scientist for Intelligent Systems, Intel, USA
Examiner is Professor Fredrik Kahl at the research group Computer vision and medical image analysis

Visual localization is a fundamental problem in computer vision, with a multitude of applications in robotics, augmented reality and structure-from-motion. The basic problem is to, based on one or more images, figure out the position and orientation of the camera which captured these images relative to some model of the environment. Current visual localization approaches typically work well when the images to be localized are captured under similar conditions compared to those captured during mapping. However, when the environment exhibits large changes in visual appearance, due to e.g. variations in weather, seasons, day-night or viewpoint, the traditional pipelines break down. The reason is that the local image features used are based on low-level pixel-intensity information, which is not invariant to these transformations: when the environment changes, this will cause a different set of keypoints to be detected, and their descriptors will be different, making the long-term visual localization problem a challenging one.  

In this thesis, five papers are included, which present work towards solving the problem of long-term visual localization. Two of the articles present ideas for how semantic information may be included to aid in the localization process: one approach relies only on the semantic information for visual localization, and the other shows how the semantics can be used to detect outlier feature correspondences. The third paper considers how the output from a monocular depth-estimation network can be utilized to extract features that are less sensitive to viewpoint changes. The fourth article is a benchmark paper, where we present three new benchmark datasets aimed at evaluating localization algorithms in the context of long-term visual localization. Lastly, the fifth article considers how to perform convolutions on spherical imagery, which in the future might be applied to learning local image features for the localization problem.

Category Thesis defence
Location: online
Starts: 22 January, 2021, 15:00
Ends: 22 January, 2021, 18:00

Page manager Published: Thu 17 Dec 2020.