Indoor Semantic Segmentation using depth information

About

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.

Camille Couprie, Clément Farabet, Laurent Najman, Yann LeCun · 2013

Related benchmarks

Task	Dataset	Result
Semantic segmentation	NYUD v2 13-class (test)	Mean Accuracy36.2	14
Semantic segmentation	NYUv2 13-class labeling	Accuracy—	12
Semantic segmentation	NYUDepth 4-Class v2	Pixel Accuracy64.5	7
Semantic segmentation	NYUDv2 4-class (test)	Mean Accuracy63.5	7
Semantic segmentation	NYUDepth 13-Class v2	Class Accuracy36.2	6