to develop abstract regions through some combination of the structure, color, and texture images
The overall goal of this work has already been established as teaching relationships among different labelled regions.Work toward this goal has included:
- the creation of (manmade) structure, color, and texture representations of images
- classifier methods that are already able to learn classes, such as a Snowman class, after lots of training data
Sal's classifier methods already work with the three different kinds of regions. The specific tasks for us now are to:
- Merge the structure regions, such that extraneous structures are filtered out.
- Create utilities for the XML files.
- Cluster the three different types of regions, (i.e., creating classes in each of those regions in hopes to use those results in creating abstract regions later).
Jocelyn: C++ programming -- multiple class organization Jenny and Clifford: webpage, database management, ground truth expansion (e.g., get pictures, label them, look into getting Corel)
Sal: make copies of the conference CD, distribute Linda's list of important articles to read Everyone: View Linda's book (focusing on chapters 6, 7, 8, and 10), learn about HMM's, EM, and support vector machines.