Recent Advances in Real-Time Road Traffic Analytics
In this presentation, I will present our latest advances on traffic surveillance. In the wake of the CVPR 2017 MIO-TCD challenge, we developed various models to analyze traffic based on ultra-low frame rate videos. This includes applications such as vehicle recognition, orientation estimation, and motion detection. We also explored solutions to compress deep conv nets on cameras whose hardware does not accommodate with large state-of-the-art convolutional neural network. The proposed compression algorithm is a budgeted dropout sparsity learning approach. We also studied ways of measuring the complexity of certain datasets following a novel "complexity measure" which allows to assess the fundamental complexity of a give image classification problem.
Pierre-Marc Jodoin is from the University of Sherbrooke, Canada where he works as a full professor since 2007. He specializes in the development of novel techniques for machine learning and deep learning applied to computer vision and medical imaging. He mostly works in video surveillance and brain and cardiac image analytics. He is the co-director of the Sherbrooke AI plateform and co-founder of the medical imaging company called "Imeka.ca" which specializes in MRI brain image analytics. His personal web site is http://info.usherbrooke.ca/pmjodoin/