Ahmad Babaeian Jelodar

Human Tracking using Particle Filtering on PETS2009 Dataset

Interactive Fountain Project in Haft-Hoz Square - Tehran

Face Recognition from Low Quality Profile

Elastic Image Registration on Face Profile Videos

Publication

Ahmad Babaeian Jelodar, Y. Sun, "Joint Object & State Recognition using Language Knowledge", IEEE International Conference on Image Processing (ICIP), 2019 (Accepted).
Ahmad Babaeian Jelodar, D. Paulius, Y. Sun, "Long Activity Video Understanding using Functional Object-Oriented Network", IEEE Transactions on Multimedia, Dec 2018, (Paper).
D. Paulius, Ahmad Babaeian Jelodar, Y. Sun, "Functional Object-Oriented Network: Construction & Expansion", IEEE International Conference on Robotics and Automation (ICRA), Brisbane, Australia, May 2018, (Paper).
Ahmad Babaeian Jelodar, M. Alizadeh, S. Khadivi, "WordNet Based Features for Predicting Brain Activity Associated with Meanings of Nouns (using fMRI images)", NAACL-HLT 2010 First Workshop on Computational Neurolinguistics, June 2010, LA, USA. (Paper)

Research and Work Projects:

Computer Vision based Safety & Security System: In this project the main goal was to create a system that can detect workers in a steel factory, and detect hazards such as wearing no helmets, identifying if a rolling machine is off or on, etc. I worked in a team of 4 Data Scientists at BCG Digital Ventures to achieve this goal. We also designed a semi-automatic labeling system for labeling human bboxes in the factory videos with classes such as person, helmet, visor, and behavioral classes such as smoking, sitting, etc.
Human tracking: In this project, a tracking by detection method is used for tracking human in consecutive frames. For detection we use HOG and DPM methods. For tracking, particle filter combined with a classifier based on adaboost for classifying detected people. All implementations were done in C/C++ using OpenCV.
Soccer analysis software: In this project, based on the tracked people in a soccer game, ball tracking and re-identification process a 2D log of the coordinates of players and the ball is constructed. The 2D log is then utilized for match reports. For this application 3 cameras are setup in a soccer stadium and we use 4 perspective points for mapping points from the frame to real world points. The match reports are extracted based on the 2D log and the analytical definitions on the log. The implementations are done in C/C++.
Conducting music using human gesture using kinect.
Interactive Fountain: In this project, a background subtraction method (ViBe in OpenCV) and some heuristic image processing of an online stream from an IP camera Installed at a fountain park in Tehran, are used to check the scene of a fountain for human movement. If humans are detected the nearest nozzle to that person is activated.
Evolutionary Algorithms (Genetic Programming) and Learning Automata at Soft Computing Laboratory, Amirkabir University of Technology: This project was implemented in Matlab.
Seperation of FSK and PSK Modulated Signals – Analyzing an Audio Signal for its Comprising Audios. (Mar. – Jun. 2011): This project was in C++. (Abstract)
Other work: Elastic Image Registration – Registration of side face video frames, Video Processing & Background Subtraction (ViBe, GMM), Biometrics – Face recognition using image filters, gabor filters, gait recognition, gait recognition.
In my current project I am using tensorflow and Python to implement state classification and a customized regularization method for deep networks.