DMCA
Learning everything about anything: Webly-supervised visual concept learning
| Venue: | In CVPR |
| Citations: | 20 - 1 self |
BibTeX
@INPROCEEDINGS{Divvala_learningeverything,
author = {Santosh K. Divvala and Ali Farhadi and Carlos Guestrin},
title = {Learning everything about anything: Webly-supervised visual concept learning},
booktitle = {In CVPR},
year = {}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
Figure 1: We introduce a fully-automated method that, given any concept, discovers an exhaustive vocabulary explaining all its appearance variations (i.e., actions, interactions, attributes, etc.), and trains full-fledged detection models for it. This figure shows a few of the many variations that our method has learned for four different classes of concepts: object (horse), scene (kitchen), event (Christmas), and action (walking). Recognition is graduating from labs to real-world ap-plications. While it is encouraging to see its potential being tapped, it brings forth a fundamental challenge to the vision researcher: scalability. How can we learn a model for any concept that exhaustively covers all its appearance varia-tions, while requiring minimal or no human supervision for compiling the vocabulary of visual variance, gathering the training images and annotations, and learning the models? In this paper, we introduce a fully-automated approach for learning extensive models for a wide range of variations (e.g. actions, interactions, attributes and beyond) within any concept. Our approach leverages vast resources of on-line books to discover the vocabulary of variance, and in-tertwines the data collection and modeling steps to alleviate the need for explicit human supervision in training the mod-els. Our approach organizes the visual knowledge about a concept in a convenient and useful way, enabling a variety of applications across vision and NLP. Our online system has been queried by users to learn models for several inter-esting concepts including breakfast, Gandhi, beautiful, etc. To date, our system has models available for over 50,000 variations within 150 concepts, and has annotated more than 10 million images with bounding boxes. 1.
Keyphrases
webly-supervised visual concept learning online system vision researcher explicit human supervision real-world ap-plications training image human supervision full-fledged detection model visual knowledge fully-automated method extensive model wide range fundamental challenge appearance variation several inter-esting concept vast resource on-line book data collection many variation different class fully-automated approach useful way visual variance appearance varia-tions
