Work

Leveraging Semantic Segmentation For Improved Image Based Localization

Year: 2022
Type: conference-paper
Abstract: In this paper we address the problem of localizing a query image in a 3D map obtained using a Structure From Motion (SfM) or a visual SLAM algorithm. Many situations, such as lighting or viewpoint changes, make the estimation process very difficult. In this paper, we have tried to increase the pose accuracy by integrating semantic information in the matching step. The classical output in semantic segmentation is a single label l for each pixel. We propose either to assign more than one label to each keypoint by two different ways. We compare the proposed methods with the state of the art. For this, we use two public datasets (Dubrovnik, Rome). We show that by incorporating visual and semantic information, the pose estimation can be improved in terms of time and precision.
Source: HAL
Institution: Institut Pascal
Language: English
Open Access status: green
OpenAlex ID: W4306858461
FWCI:
Cites:
Cited by:
Related to: 10