Designing deep networks for surface normal estimation
Authors:
- 1 June 2015
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- No. 10636919,p. 539-547
Abstract
In the past few years, convolutional neural nets (CNN) have shown incredible promise for learning visual representations. In this paper, we use CNNs for the task of predicting surface normals from a single image. But what is the right architecture? We propose to build upon the decades of hard work in 3D scene understanding to design a new CNN architecture for the task of surface normal estimation. We show that incorporating several constraints (man-made, Manhattan world) and meaningful intermediate representations (room layout, edge labels) in the architecture leads to state of the art performance on surface normal estimation. We also show that our network is quite robust and show state of the art results on other datasets as well without any fine-tuning.Keywords
All Related Versions
This publication has 19 references indexed in Scilit:
- Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation2014
- DeepPose: Human Pose Estimation via Deep Neural Networks2014
- PanoContext: A Whole-Room 3D Context Model for Panoramic Scene Understanding2014
- Perceptual Organization and Recognition of Indoor Scenes from RGB-D Images2013
- Scene Parsing by Integrating Function, Geometry and Appearance Models2013
- A category-level 3-D object dataset: Putting the Kinect to work2011
- Auto-Context and Its Application to High-Level Vision Tasks and 3D Brain Image SegmentationIEEE Transactions on Pattern Analysis and Machine Intelligence, 2009
- Recovering the spatial layout of cluttered rooms2009
- Recovering Occlusion Boundaries from a Single Image2007
- Geometric context from a single image2005