Welcome to Mohammed Hamada Gasmallah's website. I'm a machine learning researcher with a focus on computer vision and reinforcement learning.
For my undergraduate thesis project, I modified You-Only-Look-Once version 2 (YOLOv2) to predict object location and classificiation based off a series of image temporally continuous image frames. This was refined into a paper that was published in 2018 at IEMCON. Link for my published paper.
In a joint project with members of other faculty at Queen's University, we developped a computer vision technique that uses object detectors to aid in measuring multicellulor tumor spheroids of prostate cancer. This work was published at the 2019 IEEE Symposium Series on Computational Intelligence (SSCI). Here is the link for the published paper.
During the completion of my Masters, I built a computer vision machine learning system that, given an X-ray scan of a child's hand, estimates the age of the child in months. This system outperforms trained radiologists and is novel in it's use of super resolution to extract salient information from the scans. This work was published as part of the Lecture Notes in Computer Science book series. Here is a link to this work.
In order to complete the requirements for the reinforcement learning course, I completed a project implementing Multiple Models in tabular Q-learning to address the delayed reward problem.
My thesis project for my masters focused investigating the problem of object bounding box path smoothness in video object detection systems.
After the successful completion of my Masters, I decided to try to get my seminal work published at a conference. After some changes and some modifications to the work (refining experiments and analysis), This paper was successfully published as part of The 36th Canadian Conference on Artificial Intelligence.
I was inspired by the NeRF and NGP papers to attempt to recreate them using Jax. This project was my attempt at doing this. All images generated are at a relatively low resolution (100x100).
This is a game that I designed and developped with a team of four for Ludum Dare 38. Ellipticats is a 2D physics-based platformer in which you search the galaxy for lost cats. You play as a rover sent by the cat homeworld to retrieve citizens lost in a cat-astrophe of astronomical proportions. Use your wheels and thrusters to hop between planets and reunite a community. This game placed 19th in terms of fun!
If you'd like to get in touch for hiring opportunities, consulting opportunities, to discuss research opportunities or just to have a chat, this it the way to get in touch with me!