← Back to papers
2015 ICCV ICCV 2015

Single Image 3D Without a Single 3D Image

Abstract

Do we really need 3D labels in order to learn how to predict 3D? In this paper, we show that one can learn a mapping from appearance to 3D properties without ever seeing a single explicit 3D label. Rather than use explicit supervision, we use the regularity of indoor scenes to learn the mapping in a completely unsupervised manner. We demonstrate this on both a standard 3D scene understanding dataset as well as Internet images for which 3D is unavailable, precluding supervised learning. Despite never seeing a 3D label, our method produces competitive results.

🌉 Interdisciplinary Bridge — Computer Vision and Deep Learning
📈 Trend Setter — Self-Supervised Learning
🐣 Hot Topic Early Bird — self-supervised learning
🐝 Cross-Pollinator — Artificial Intelligence, Computer Science, Computer Vision, Data Science & Analytics, Deep Learning, Healthcare & Medicine, Interdisciplinary, Knowledge & Reasoning, Machine Learning, Mathematics & Optimization, Natural Language Processing, Reinforcement Learning, Robotics, Security & Privacy, Speech & Audio

Authors

David F. Fouhey , Wajahat Hussain , Abhinav Gupta , Martial Hebert

Topics

Computer Vision > Analysis > 3D Vision Computer Vision > Analysis > Depth Estimation Computer Vision > Analysis > Scene Understanding Deep Learning > Learning Types > Self-Supervised Learning

Keywords

unsupervised learning representation learning 3d reconstruction scene understanding self-supervised learning depth estimation indoor scene
Download PDF

Related papers

Cutting Edge: Soft Correspondences in Multimodal Scene Parsing 2015
Unsupervised Generation of a Viewpoint Annotated Car Dataset From Videos 2015
Depth-Based Hand Pose Estimation: Data, Methods, and Challenges 2015
Peeking Template Matching for Depth Extension 2015
Learning Semi-Supervised Representation Towards a Unified Optimization Framework for Semi-Supervised Learning 2015