Experience

• 

  Sharif university of technology

  Sept 2007 - Jan 2010

  Research assistant

  Research on image processing, computer vision, and machine learning.Teaching

• 

  Azad university

  Jan 2010 - Jun 2010

  Lecturer

  Microprocessors

• 

  Idiap research institute

  Sept 2010 - Jan 2011

  Trainee

  - Scene RecognitionThe goal of the project was to help a robot recognize and navigate in different types of rooms using visual and depth cues. We developed a machine-learning system that employs SVM and convolutional neural networks (a deep-learning approach) in this task. I implemented these algorithms in MATLAB and Python. Our studies show that a combination of color and depth information leads to highest recognition accuracies.- Multimodal Object ClassificationIn this project, I developed a system that detects objects by combining visual cues and the voice commands received from the user, using MFCC features for the voice module and SIFT features for the visual data. I compared combining data at feature level, basis vector level, and decision level. The results of my study show that data fusion at the decision level leads to the best classification performance. Weniger anzeigen

• 

  Universität bern

  Aug 2012 - Jan 2016

  Research assistant

  In the total knee replacement surgery, the diseased parts of the bones are trimmed and replaced with implants. In this surgery the best possible treatment strategy varies from patient to patient. Yet, there was no technique available to help the surgeon decide on the surgery plan based on patient information. The goal of my thesis was to devise a fully automatic system that predicts the outcome of surgery based on pre-operative images. To design such a system, I used different image processing and machine learning techniques such as image registration, statistical shape modeling, and the random forest algorithm. Weniger anzeigen

• 

  Institute of social and preventive medicine (ispm)

  Jun 2016 - Feb 2018

  Postdoctoral researcher

  I am working on predicting the healthy shape of the patient's bone before it has been affected by an orthopedic disease such as osteoarthritis. The outcome of this prediction helps the surgeon in recovering the geometry of the healthy bone. To solve this problem, I am building a statistical model representing the shape variations of healthy bones in the population. The model is then fitted to the healthy portion of the patient's bone to predict the missing parts affected by the disease. So far, I proposed and implemented a fully automatic segmentation technique that finds the anatomical correspondences among bones, an essential step in fitting the population model to the patient's bone. Additionally I developed a software with a graphical interface that allows the user to load the image and produces the segmented bone. Weniger anzeigen

• 

  Fraunhofer mevis

  Jan 2018 - now

  Research scientist