RF using Machine Learning Algorithms:

Our test database contains more than 20,000 general-purpose images from Corel Image Gallery. They cover a variety of topics, such as “flower”, “tiger”, “eagle”, “gun”, “horse”, etc. We use the color correlogram as the low-level feature. To collect user feedback log, we invite 8 subjects to query images using our system. Totally, we accumulate 1,500 query sessions.

· A RF Example using Bayesian Learning

· A RF Example using SVM

· A RF Example using ML Framework

Retrieval Performance Comparison:

Query by Keywords:

In the first example, we use the keyword "flower" to query images. In the second example, the keyword for query is "model".

· Annotation Propagation by Bayesian Learning

· Annotation Propagation by SVM

· Annotation Propagation by ML Framework