On April 19, 2013, Nicolae Suditu successfully defended his PhD thesis entitled "Adaptive relevance feedback for large-scale image retrieval".
This work investigates several extensions of a state-of-the-art content-based image retrieval system. The two main contributions are an algorithmic extension to make it suitable for collections of one million images or more, and a novel procedure to smoothly transition from an exploratory phase to an exploitation one.
This research was funded by the Hasler Foundation, under grants EMMA and UBSL.
The resulting algorithms are freely available under the AGPL3.0 open-source license.
For more details click here: Adaptive relevance feedback for large-scale image retrieval