Dr. Reza Bahmanyar

I develop algorithms based on machine learning (e.g., generative Bayesian models and Deep learning) for extracting meaningful information from Earth Observation image time series. These algorithms will be used in an ecosystem monitoring system.

I developed new approaches based on machine learning and statistics to extract meaningful information from Earth Observation images. I employed various methods such as Support Vector Machines (SVM), logistic regression, and latent Dirichlet allocation (LDA) for supervised and unsupervised data mining and classification. Further, I proposed novel feature selection and evaluation methods for image mining systems. My developed approaches will be used for data mining of large-scale image archives.

I developed a semi-supervised and model-based method for building extraction and 3D model reconstruction based on optical satellite images and height information.

I improved automatic discovering of human-understandable object attributes from images and text data. I proposed new approaches based on LDA, spectral clustering, and SVM methods. This improvement scales the object recognition systems applicability to the large volume of images and text data available on the online shopping web sites. It also increases the relevancy of the search results to the user queries.

I guided the Computer Graphics students of Prof. Philipp Slusallek to build a ray tracing system or hardware-based 3D visualization application in C++ programming language.

I assisted in installing and calibrating the Point Grey's video cameras in the human activity recognition lab. Moreover, I operated the Point Grey's Bumblebee cameras for stereo vision tasks.

I assisted in calibrating sensors and transmitters and installing industrial analyzers.

Machine learning, Data mining, Computer vision, Pattern recognition, Image processing

Computer vision, Image processing, Computer graphics

Electronics, Optimization algorithms