Michael Iliadis

I am a Research Engineer at Vidado.ai. My current focus is in document image analysis (form recognition and key-value object detection). My broader research interests include deep learning and sparse modeling for image processing and computer vision applications.

Prior to Vidado, I was a research scientist at SONY US Research Center where I worked on semantic segmentation for SONY products.

I received my Ph.D. in EECS from Northwestern University in 2016 where I worked under the supervision of Aggelos K. Katsaggelos in the Image and Video Laboratory (IVPL). I received my M.S. degree in Computer Science from the University of Bath, UK in 2009 and the B.S. degree in Digital Systems from the University of Piraeus, Greece in 2008.

Email  /  CV  /  Thesis  /  Google Scholar  /  LinkedIn

Research

My research focuses in video compressive sensing for perceived quality video reconstruction and recognition. The methodologies and techniques I have applied include sparsity-seeking optimization and deep learning based models. In addition, I have delivered and continue to work on real-world problems such as semantic segmentation for video scene understanding and image retrieval.

DeepBinaryMask: Learning a Binary Mask for Video Compressive Sensing
Michael Iliadis, Leonidas Spinoulas, Aggelos K. Katsaggelos
Elsevier Digital Signal Processing, 2020
bibtex / code & data

We propose a novel encoder-decoder neural network model called DeepBinaryMask for video compressive sensing. The proposed framework is an end-to-end model where the sensing matrix is trained along with the video reconstruction.

Using Deep Neural Networks for Inverse Problems in Imaging
Alice Lucas, Michael Iliadis, Rafael Molina, Aggelos K. Katsaggelos
IEEE Signal Processing Magazine (SPM), 2018
bibtex

We review the popular neural network architectures used for imaging tasks, offering some insight as to how these deep-learning tools can solve the inverse problem.

Deep Fully-Connected Networks for Video Compressive Sensing
Michael Iliadis, Leonidas Spinoulas, Aggelos K. Katsaggelos
Elsevier Digital Signal Processing, 2018
project page / bibtex / code & data / supplement

A deep learning framework for video compressive sensing.

Robust and Low-Rank Representation for Fast Face Identification with Occlusions
Michael Iliadis, Haohong Wang, Rafael Molina, Aggelos K. Katsaggelos
IEEE Transactions on Image Processing (TIP), 2017
bibtex / code

A fast iterative method to address the face identification problem with block occlusions.

Multi-Model Robust Error Correction for Face Recognition
Michael Iliadis, Leonidas Spinoulas, Albert S. Berahas, Haohong Wang, Aggelos K. Katsaggelos
International Conference Image Processing (ICIP), 2016
bibtex

The proposed formulation allows the simultaneous use of various loss functions for modeling the residual in face images.

Block Based Video Alignment with Linear time and Space Complexity
Armin Kappeler, Michael Iliadis, Haohong Wang, Aggelos K. Katsaggelos
International Conference Image Processing (ICIP), 2016
bibtex

We propose a fast, robust and memory efficient video sequence alignment algorithm which has linear space and time complexity.

Sparse Representation and Least Squares-based Classification in Face Recognition
Michael Iliadis, Leonidas Spinoulas, Albert S. Berahas, Haohong Wang, Aggelos K. Katsaggelos
European Signal Processing Conference (EUSIPCO), 2014
bibtex / code

Effectively, our method combines the sparsity-based approaches with additional least-squares steps.

Virtual touring - A Content Based Image Retrieval application
Michael Iliadis, Seunghwan Yoo, Xin Xin, Aggelos K. Katsaggelos
International Conference on Multimedia and Expo Workshops (ICMEW), 2013
bibtex

A content based image retrieval application for searching landmarks and buildings in a city using a smartphone.

Video Compressive Sensing using Multiple Measurement Vectors
Michael Iliadis, Jeremy Watt, Leonidas Spinoulas, Aggelos K. Katsaggelos
International Conference Image Processing (ICIP), 2013
bibtex / Top 10% Paper Recognition

The approach takes advantage of Multiple Measurement Vectors (MMV), seeking for significantly sparser solutions, assuming that the solution vectors have similar sparsity structure.

Teaching

EECS 214: Data Structures and Data Management - Spring 2015, 2016

GEN_ENG 205: Engineering Analysis - Winter 2015


I copied the source code of this website from here.