Alert icon
We're changing our privacy policy. This stuff matters.  Learn more  Dismiss

Large-scale Image Classification: ImageNet and ObjectBank

Loading...

Sign in or sign up now!
Alert icon
Upgrade to the latest Flash Player for improved playback performance. Upgrade now or more info.
3,376
Loading...
Alert icon
Sign in or sign up now!
Alert icon
There is no Interactive Transcript.

Uploaded by on May 18, 2011

Google Tech Talk (more info below)
May 5, 2011

Presented by Professor Fei-Fei Li, Stanford University

ABSTRACT

A key challenge in visual recognition is to recognize and label a large number of visual concepts, such as object and scene classes. In this talk, I'll discuss two recent projects in the Stanford Vision Lab on this topic. ImageNet is a large-scale image ontology that is built on the backbone structure of WordNet. In this talk, we show briefly how ImageNet is constructed using Amazon Mechanical Turk. And given the largest (publicly available) dataset of tens of thousands of image classes, how today's state-of-the-art computer vision algorithms do in the problem of large-scale image classification? Then I will discuss a new image representation called "Object Bank", that is a significant departure from all the previous image representation techniques such as Bag-of-Words models using low-level features (SIFT, HOG, GIST, etc.). We show that using the new Object Bank representation, a simple linear SVM classifier can result in superior performances in all standard image classification datasets. Furthermore, sparsity algorithms make our representation more efficient and scalable for large scene datasets, and reveal semantically meaningful feature patterns.

  • likes, 4 dislikes

Link to this comment:

Share to:
see all

All Comments (16)

Sign In or Sign Up now to post a comment!
  • very nice video.

  • cool vedio

  • great vodeo.

  • @lennyhome I give you the 'Troll of the vid.' award. Now take the price and continue to pollute Internet in other place.

  • @lennyhome You have serious psychological problems. But if you think you are winner, lets be it. By the way, if people stop talking to you, it is not because they lack arguments. It is more because they don't want to spent time with you :)

  • @lennyhome The title and the abstract are more than enough for a person of say moderate intelligence to get it.

  • @mercatormac What's a sarcasm? Also the problem here is that I don't have an account on ieeexplore and I'm not going to make one so, if you can summarize in a comment the reason why you're spamming that address we can continue our little discussion, otherwise I'm done and the winner.

  • @lennyhome ieeexplore(dot)ieee(dot)org/xp­l/freeabs_all(dot)jsp?arnumber­=5485792

    Sarcasm

    How do these bastards sleep at night?

    /Sarcasm

    Hope this helps.

  • @mercatormac I don't know. Maybe they take pills. You tell me. You're the smart one here.

  • @lennyhome ieeexplore(dot)ieee(dot)org/xp­l/freeabs_all(dot)jsp?arnumber­=5485792

    How do these bastards sleep at night?

    PS You're an idiot.

Loading...

Alert icon
0 / 00Unsaved Playlist Return to active list
    1. Your queue is empty. Add videos to your queue using this button:
      or sign in to load a different list.
    Loading...Loading...Saving...
    • Clear all videos from this list
    • Learn more