Sasan Asadiabadi, Ph.D.

rtMRI-VTAT is a Python-based GUI and a vocal tract analysis tool for rtMRI datasets.

My Ph.D research was centered around localizing anatomical landmarks of the vocal tract in real-time MRI videos and analysing how speech articulators change over time as we utter a sentence. VTAT is a GUI-based vocal tract analysis tool, which aims to help the researchers to track vocal tract contours in MRI videos and study the fundementals of human speech production. VTAT is based on a light-weight MobileNet-V3 backbone and a UNet architecutre, which estimates the vocal tract landmarks via heatmap regression.

Smart AI for early breast cancer detection and diagnosis.

In my role as an AI engineer, I conducted extensive research around design, training, optimization, evaluation and deployment of deep learning algorithms for detection of breast cancer in 2D and 3D mammograms, aimed at improving performance to the level of expert radiologists and beyond.

Conducting research project under the [4TU Precision Medicine](https://www.4tu.nl/precision-medicine/).

I conduct my postdoctoral research as a member of Laboratory of Biophysics BIP at WUR. My research mainly focuses on applications of deep learning in analysis of quantitative MRI. I implement neural networks for spatio-temporal modeling of 3D MRI data and unsupervised clustering of biological tissues.

Responsible for development of intelligent inspection systems for safer cities.

RaSpect is a drone-based intelligent inspection company and over there I did research and development of deep learning algorithms centered around the facade element and defect detection. We utilized drones to capture images/videos of facades and developed deep models to localize the deffective areas in the orthophoto stiched images. Check our patent at Systems and methods for artificial intelligence powered inspections and predictive analyses

Responsible for development of real-time driver monitoring system.

In my role as a computer vision engineer at CY Vision, I did extensive research on deep learning on edge. I developed efficient light-weight neural networks for tracking driver’s facial landmarks and pupils from a stero-vision camera behind the steering wheel, robust to occlusion and different lighting conditions. I worked on optimizing and deploying the deep learning models on edge devices like low power ARM processors with C++.

In-cabin driver 3D gaze and pose estimation system using monocular vision.

At Eyeris I developed a real-time 3D gaze and pose estimation system for tracking driver’s behavior in the video stream being captured from the central mirror.

I conducted research in the topic of applications of computer vision and deep learning in medical image analysis.

During my M.Sc. and Ph.D. career, I have mainly focused on analyzing medical images using various deep learning and computer vision techniques. I have worked on several projects including localizing anatomical landmarks on human vocal tract in real time MRI sequences, biomedical image segmentation, blind estimation of vocal tract articulatory features from speech and generating visual face animations synchronized with input speech.

Teaching a Python programming language cource.

In my academic career, I have gained teaching experience by fulfilling my teaching assistance duties for several undergraduate and graduate courses. As well, I have lectured a course “introductionto programming with Python” to undergraduate students with non engineering backgrounds.

Under supervision of Prof. Engin Erzin.

Thesis title: Deep Learning Approaches for Vocal Tract Boundary Segmentation in rtMRI,

Under supervision of Prof. Engin Erzin.

Thesis title: Vocal Tract Contour Tracking for real-time Speech MRI Using a Shape Model Prior.

Under supervision of Dr. Gholampour.

Project title: Voice Activity Detection for Squelch Systems in Noisy Environment.