Abstract : This work addresses multi-class segmentation of indoor scenes with RGB-D inputs. While this area of research has gained much attention recently, most works still rely on hand-crafted features. In contrast, we apply a multiscale convolutional network to learn features directly from the images and the depth information. We obtain state-of-the-art on the NYU-v2 depth dataset with an accuracy of 64.5%. We illustrate the labeling of indoor scenes in videos sequences that could be processed in real-time using appropriate hardware such as an FPGA.
Type de document :
Communication dans un congrès
First International Conference on Learning Representations (ICLR 2013), May 2013, Scottsdale, AZ, United States. pp.1-8, 2013
https://hal.archives-ouvertes.fr/hal-00805105
Contributeur : Laurent Najman <>
Soumis le : mercredi 27 mars 2013 - 09:33:18
Dernière modification le : mercredi 27 juillet 2016 - 14:48:48
Camille Couprie, Clément Farabet, Laurent Najman, Yann Lecun. Indoor Semantic Segmentation using depth information. First International Conference on Learning Representations (ICLR 2013), May 2013, Scottsdale, AZ, United States. pp.1-8, 2013. <hal-00805105>
