What Does Classifying More Than 10, 000 Image Categories Tell Us?

نویسندگان

  • Jia Deng
  • Alexander C. Berg
  • Kai Li
  • Li Fei-Fei
چکیده

Image classification is a critical task for both humans and computers. One of the challenges lies in the large scale of the semantic space. In particular, humans can recognize tens of thousands of object classes and scenes. No computer vision algorithm today has been tested at this scale. This paper presents a study of large scale categorization including a series of challenging experiments on classification with more than 10, 000 image classes. We find that a) computational issues become crucial in algorithm design; b) conventional wisdom from a couple of hundred image categories on relative performance of different classifiers does not necessarily hold when the number of categories increases; c) there is a surprisingly strong relationship between the structure of WordNet (developed for studying language) and the difficulty of visual categorization; d) classification can be improved by exploiting the semantic hierarchy. Toward the future goal of developing automatic vision algorithms to recognize tens of thousands or even millions of image categories, we make a series of observations and arguments about dataset scale, category density, and image hierarchy.

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

منابع مشابه

Searching in document images: what does the appearance of a document tell us about what it means?

The document understanding problem can be informally defined as the automatic extraction of meaning from documents. In the Intelligent Sensory Information Systems group we have experimented with analyzing the visual appearance of documents in order to extract meaning. That is, we concentrate on how documents look, rather than on what they say. We motivate this approach with several applications...

متن کامل

What does the residual income method tell us about housing affordability in Australia?

This bulletin is based on research conducted by Professor Terry Burke and Ms Liss Ralston at the AHURI UNSWUWS Research Centre, and Professor Michael Stone from the University of Massachusetts, USA. The project considered the policy implications of using a different measure for housing affordability— using the residual income rather than the income ratio method. What does the residual income me...

متن کامل

The Natural Rights of Children

What does libertarian theory, Murray Rothbard’s theory in particular, tell us about the rights of children? The two foundational principles of Rothbardian libertarianism are the sanctity of private property and the rule of non-aggression. Persons, including children, are “self-owners”. Yet children, at a young age, are not yet capable of functioning fully as “self-owners.” They must be cared fo...

متن کامل

Topics in Representation Theory: Maximal Tori and the Weyl Group

where the Hilbert space direct sum is over all irreducible representations of G. Under the left action of G, the infinite dimensional representation on L(G) breaks up into finite dimensional representation spaces Vπ with multiplicity dim Vπ. What the Peter-Weyl theorem doesn’t do is tell us what the representations are and how to tell them apart. We would like to know how to project from L(G) o...

متن کامل

Facts, Events, Things and the Ontology of Physics

The philosophical importance of notions like ‘facts’, ‘events’, and ‘things’ lies in the central role they play in ontology, the venerable branch of metaphysics which deals with the question of what there is. Depending on various philosophical options, reality is regarded as either constituted by things, or by events, or by facts, or by a pluralistic combination of these categories. But what, p...

متن کامل

ذخیره در منابع من


  با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید

برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید

ثبت نام

اگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید

عنوان ژورنال:

دوره   شماره 

صفحات  -

تاریخ انتشار 2010