Members of the PSI Lab study machine learning, genome biology and vision. We develop statistical inference and probabilistic reasoning (graphical modeling) methods for learning about complex patterns in data. In our genome biology research, we are interested in understanding the regulatory processes that enable a limited number of genes to generate a much more massive and diverse set of genetic messages. In our vision research, we focus on developing probabilistic generative models for scene and object analysis. The group is led by Brendan J. Frey in the Department of Electrical and Computer Engineering, with cross appointments in Computer Science, Banting and Best Department of Medical Research and Donnelly Centre for Cellular and Biomolecular Research. If you are interested in joining the group, click here.
Current Research Highlight
A new approach to vision: Flobject analysis
Unsupervised learning can be used to extract image representations that are useful for various and diverse vision tasks. After noticing that most biological vision systems for interpreting static images are trained using disparity information,
we developed an analogous framework for unsupervised learning. The output of our method is a model that can generate a vector representation or descriptor from any static image. However, the model is trained using pairs of consecutive video frames, which are used to find representations that are consistent with optical flow-derived objects, or ‘flobjects’. To demonstrate the flobject analysis framework, we extend the latent Dirichlet allocation bag-of-words model to account for real-valued word-specific flow vectors and image-specific probabilistic associations between flow clusters and topics. We show that the static image representations extracted using our method can be used to achieve higher classification rates and better generalization than standard topic models, spatial pyramid matching and gist descriptors.
Reference
PS Li, IE Givoni, BJ Frey, Learning better image representations using flobject analysis. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2011. [pdf file].