Real-time 3D Scene Layout from a Single Image Using

Convolutional Neural Networks


   Shichao Yang, Daneil Maturana, Sebastian Scherer

   shichaoy@andrew.cmu.edu

Pop up

Abstract

We consider the problem of understanding the 3D layout of indoor corridor scenes from a single image in real time. Identifying obstacles such as walls is essential for robot navigation, but also challenging due to the diversity in structure, appearance and illumination of real-world corridor scenes. Many current single-image methods make Manhattan-world assumptions, and break down in environments that do not meet this mold. They also may require complicated hand-designed features for image segmentation or clear boundaries to form certain building models. In addition, most cannot run in real time. In this paper, we propose to combine machine learning with geometric modelling to build a simplified 3D model from a single image. We first employ a supervised Convolutional Neural Network (CNN) to provide a dense, but coarse, geometric class labelling of the scene. We then refine this labelling with a fully connected Conditional Random Field (CRF). Finally, we fit line segments along wall-ground boundaries and pop up a 3D model using geometric constraints. We assemble a dataset of 967 labelled corridor images. Our experiments on this dataset and another publicly available dataset show our method outperforms other single image scene understanding methods in pixelwise accuracy while labelling images at over 15Hz.

Video



Dataset

(Download)

This is an image dataset (967 images) for corridor environments. All the images are annotated as ground or wall using polygons. It is a mixture of processed SUN RGBD (349, corridor category), SUN Database (327, corridor category), and self-collected images (291, in CMU). If you use this dataset, you should also cite two other paper listed in the bibtex.


dataset

Paper

(PDF)

Bibtex

[Bibtex]
Last updated: Sept 17, 2016
StatCounter - Free Web Tracker and Counter