Improving semantic segmentation in urban scenes with a cartographic information

Abstract : This paper presents three different approaches to inject a location information in semantic segmentation Convo-lutional Neural Networks (CNN) applied to urban scenes. The assumption that a location information would improve semantic segmentation performance emerges from the idea that some elements of urban scenes are located in a predictable manner. This assumption is confronted to realistic data on the CARLA autonomous driving simulator, which is used to create our own synthetic dataset with images, depth maps and bird-eye-view cartographic images. Simulators circumvent the difficulties due to the scarcity of publicly available synchronous labeled images and location information. We consider the location information as a cartographic image as we assume it is the simplest option to include it in a CNN. We assess the relevance of injecting the cartographic information in three different manners: as a CRF potential, as an additional task and as an additional encoder input of a CNN. The three methods are evaluated and compared with a state of the art CNN with regards to the pixel-wise accuracy, mean intersection over union and intersection over union of some important classes. The multi-encoder approach improves the intersection over union of the pedestrians, vehicles and traffic signs classes by respectively 4%, 1.6% and 9 %.
Type de document :
Communication dans un congrès
The 15th International Conference on Control, Automation, Robotics and Vision (ICARCV 2018), Nov 2018, Singapore, Singapore
Liste complète des métadonnées

https://hal.archives-ouvertes.fr/hal-01875096
Contributeur : Vincent Fremont <>
Soumis le : dimanche 16 septembre 2018 - 16:02:09
Dernière modification le : lundi 12 novembre 2018 - 11:10:08

Fichier

icarcv-final-submission.pdf
Fichiers produits par l'(les) auteur(s)

Identifiants

  • HAL Id : hal-01875096, version 1

Citation

Abdelhak Loukkal, Vincent Frémont, Yves Grandvalet, You Li. Improving semantic segmentation in urban scenes with a cartographic information. The 15th International Conference on Control, Automation, Robotics and Vision (ICARCV 2018), Nov 2018, Singapore, Singapore. 〈hal-01875096〉

Partager

Métriques

Consultations de la notice

45

Téléchargements de fichiers

41