Hossein Rahmani
Dr. Hossein Rahmani is currently an Assistant Professor (Lecturer) with the School of Computing and Communications, Lancaster University. He was a Research Fellow in the School of Computer Science and Software Engineering (CSSE), the University of Western Australia (UWA), Perth. He obtained his Ph.D. in Computer Vision and Machine Learning from UWA. He has developed machine learning systems and various feature extraction algorithms for RGB-Depth Action Recognition. He has published research papers in high impact factor journals and reputable conferences.
News
*Sep. 2018: A paper has been accepted in ACCV'18 for Oral presentation​​
*Aug. 2018: Our Journal paper titled "Identity Adaptation for Person Re-identification" has been published in IEEE Access
*May 2018: Our journal paper titled "Three-Dimensional Scanning for Measurement of Bulk Density in Gravelly Soil" has been published in Soil Use and Management
*Feb. 2018: Our book titled "A Guide to CNNs for Computer Vision" has been published by Morgan & Claypool
*Oct. 2017: Our application to the UWA Research Collaboration Awards has been accepted. Project Title: "Advanced Robotics with 3D Computer Vision for Automatic Healthcare Monitoring," (amount: AU$27,845)
*Aug. 2017: ICCV'17 Young Researcher Awards which support to attend ICCV'17
*July 2017: I delivered a part of the "Deep Learning for Computer Vision" course, at the International Summer School on Deep Learning (DeepLearn2017) in Bilbao, Spain. (Slides)
*July 2017: My paper on Action Recognition accepted in ICCV 2017
Hossein Rahmani and Mohammed Bennamoun, "Learning Action Recognition Model from Depth and Skeleton Videos", ICCV'17 (PDF)
*Mar. 2017: The following paper has been accepted in TPAMI 2017
Hossein Rahmani, Ajmal Mian and Mubarak Shah, "Learning a Deep Model for Human Action Recognition from Novel Viewpoints", Link to IEEE version
MY LATEST RESEARCH
Human Pose Model: CNN + temporal modeling for action recognition (CVPR 16 Oral)
CNN is trained using synthetic 3D human models combined with real mocap data to learn view invariant representation. Group Sparse Fourier Temporal Pyramid on the CNN features is used for classification and dramatic improvement over state-of-the-art is achieved.
NKTM: Nonlinear Knowledge Transfer Model for cross-view action recognition (CVPR 15)
NKTM is a deep network, with weight decay and sparsity constraints, which finds a single shared high-level virtual path from videos captured from different unknown viewpoints to the same canonical view. A single NKTM is learned for all actions and all camera viewing directions without action labels.
HOPC for cross view action recognition (PAMI 2016)
Histogram of Oriented Principal Component features are extracted at local spatio-temporal regions after view-normalization. These features are then combined with Spatio-temporal Keypoint Distribution features to perform view-invariant human action recognition.