About me

This is Javad Zolfaghari-Bengar’s home on the web!

I received Ph.D. from Computer Vision Center(CVC) of Universidad Autonoma de Barcelona with a thesis on Reducing Label Effort with Deep Active Learning in LAMP team, advised by Dr. Joost van de Weijer and Dr. Bogdan Raducanu in December 2021.

My doctoral research revolved around active learning to reduce label effort in deep learning methods. In the context of Autonomous driving, we introduced Temporal Coherence method for object detection in videos. Also addressed the problem of biased sampling in imbalanced datasets using an optimization-based method. Moreover, we proposed label-dispersion metric based on learning dynamics of neural networks to alleviate the over-confidence in neural nets. Eventually we studied the incorporation of self-supervised learning in active learning to reduce label effort. My main research interests include deep neural nets, object detection, instance segmentation, Active Learning, Transfer Learning and Domain Adaptation.

Previously I completed M.Sc. degree from Universitat Politecnica de Catalonia in Information and Communications Technology with the thesis on Radio Resource Management in OFDMA Networks that I did in the Institute of Information Theory of RWTH-Aachen University in 2014.