PhD Proposal: Rich and Efficient Visual Data Representation

Talk
Mohammad Rastegari
Time: 
04.24.2014 15:00 to 16:30
Location: 

AVW 3258

Increasing the size of training data in many computer vision tasks has shown to be very effective. Using large scale image datasets (e.g. ImageNet) with simple learning techniques (e.g. linear classifiers) one can achieve state-of-the-art performance in object recognition compare to sophisticated learning technique on smaller image sets. Semantic search on visual data become very popular. There are billions of images on the internet which are increasing every day. Dealing with large scale image sets is intense per se. They take heavy amount of memory that makes it impossible to process the images with complex algorithms on single CPU machines. Finding an efficient image representation can be a key to detract this problem. A representation to be efficient is not enough for image understanding. It should be comprehensive and rich in carrying semantic information. In this proposal we show binary codes as a rich and efficient image representation. We demonstrate several tasks in which binary features can be very effective. We show how binary features can speed up the large scale image classification. We present learning techniques to learn the binary features from supervised image set (With different ways of semantic supervision; class labels, textual descriptions). We propose several problems that are very important in finding and using efficient image representation.
Examining Committee:
Committee Chair: - Dr. Larry Davis
Dept. Representative - Dr. Hal Daume III
Committee Member - Dr. David Jacobs