Mine4Nuggets. Searching for objects into images
|
Supervisor |
(James Martin Fellow (http://www.oxfordmartin.ox.ac.uk/people/236), Fellow of the Oxford Man Institute James Martin Fellow
(http://www.oxfordmartin.ox.ac.uk/people/236), Fellow of the Oxford Man Institute)
|
|
Suitable for |
MSc in Computer Science
|
Abstract
(Joint supervisior with G Orsi) Most of the information on commercial websites (e.g., Amazon, Rightmove, and Autotrader) is rep-resented by (semi-)structured pages where the interesting data is structured into de_ned structured or visual patterns. However, some critical information is often represented with an image or a chart. As an example, on many real-estate websites, data about the energy e_ciency of the property is represented by an EPC Chart (http://www.epcchart.com/), while the actual size of the rooms is represented through a oor-plan http://en.wikipedia.org/wiki/Floor_plan). Have you ever tried to search for an energy-e_cient property or one with a large bedroom? The answer is: you can't. This project is concerned with the study of semantic image analysis techniques that can be used to locate and extract interesting objects from pictures and charts on the web. The goal is to enhance current search technology with object-search capabilities for images, a _eld that o_ers challenging research and potential commercial opportunities. The project will be conducted within the DIADEM project (http://diadem.cs.ox.ac.uk). Good knowledge of Java is required, previous knowledge about image analysis, computer vision and OCR technology is preferred.