Image of Hassan Askary

Hassan Askary

I work with images.I write about computer vision here and you can see my photography here.


Jun - Jul 2023

Computer Vision Research Intern in the Optical Engineering team at Cognex where I investigated deep learning methods to expand the depth of field of machine vision cameras with big apertures.


2022 - Present

Awarded the Erasmus Mundus Joint Master's Degree scholarship to study in the Computational Colour and Spectral Imaging program at NTNU, Norway and UGR, Spain.


2021 - 2022

Associate Data Scientist in the AI team at i2c Inc. where I delivered computer vision based products like the automated cheque deposit service which uses computer vision and deep learning algorithms to read and parse handwritten information on bank cheques.


2017 - 2021

BSc at PIEAS in Computer Science during this I worked at Techlets as a Data Science intern building NLP based applications for automating extraction of geolocation information from text. I also worked as a teaching assistant for the computer vision course.


Projects


Identity Based Face Synthesis


Developed method to control extraction of specific data points from the latent space of GANs. Proposed CGAN + Autoencoder for generating IDs from faces and faces from corresponding IDs. Deployed model on Nvidia Jetson Nano and a Flask web app.Stack used: PyTorch, Nvidia CUDA, Flask, Arduino


Survey of Deep Learning Methods for Single Image Super Resolution


Literature review of deep learning methods to achieve single image super resolution. Identified the current state-of-the-art method. Developed taxonomy and compared performance. Identified approaches that deliver results. Proposed future areas of research and limitations of current methods


Semantic Appearance Manipulator


Image manipulation program using Mask R-CNN. Replaces detected objects in image with another image. Developed using PyTorch, Numpy, Matplotlib, and Pillow


Creation of Small Footprint FOGRA 39 CMYK ICC Profile


Developed a small footprint ICC profile creation method. Reduced size of Fogra 39 CMYK profile by 74%. Wrote Python script to automate small footprint profile creation. Wrote Python script to convert ICC profiles from v2 to v4.