PhD in Computer Science, 2019
Aston University, UK
MSc in High Speed Mobile Networks, 2012
Oxford Brookes University, UK
BSc in Telecommunication and Networks, 2010
COMSATS University, Pakistan
Aamir AKbar is the co-founder of the AWKUM AI lab. His research is inspired by natural processes, and engineering of resource-constrained intelligent systems. His most notable achievements are in the area of Artificial Intelligence. Specifically, his work leverages multi-objective optimization, evolutionary computation, machine learning, self-adaptivity and self-awareness to tackle problems, and has been applied in areas such as Cyber-Physical Systems, Cloud Computing, Internet-of-Things, and Software Defined Networks. His work has been published in some of the prestigious IEEE journals and Transactions. In addition, his research has been featured in global press, including the BBC (UK), Daily Mail (UK), Hindustan Times (India), Xinhua(China) and The News (Pakistan).
I'm currently looking for MS/PhD level students to work on various research projects, some of which are listed below.
Memes on social media are frequently amusing, but with the appropriate image and wording, they can quickly become offensive. This project aims to create a multimodal to comprehend both the picture and the text's meanings, as well as how the image influences the text's goals, and to determine if the meme is hateful or not. Strong programming skills (ideally in Python), an interest in NLP, and a CV are required. Basic understanding or expertise with deep learning is a bonus.
The goal of this research is to investigate if Deep Reinforcement Learning can be utilized to create an end-to-end algorithm that can assist blind and visually impaired people in navigating the streets. On their walks, a smartphone application keeps them on the sidewalk, guides them to pedestrian crossings and past barriers, and detects the condition of the pedestrian traffic light, ensuring that they arrive safely at their destination. Strong Python programming skills (TensorFlow/Pytorch) is required.
This project's aim is to investigate what the best auto-encoder (AE) architecture is for encoding geometrical shape. Statistical shape models have been the gold standard for modeling normal shape variation for downstream applications like image segmentation for decades. Deep generative models have recently been used to build shape priors that are used to regularize tasks such as image segmentation, although it is still unclear how they (optimally) encode shape. Strong Python programming skills (TensorFlow/Pytorch) is required.
The purpose of this research is to develop deep learning models for the task of summarizing scientific papers. The number of scientific articles released is continually rising, making it nearly difficult for researchers to keep up with the most current findings. Automatic text summary for scientific articles is one logical solution to this challenge. On large-scale datasets of scientific paper summarization, we aimed to explore deep learning algorithms. Python programming skills and strong interest in natural language processing are required.
In recent years, more experts have begun to pay attention to news picture captioning. It connects the dots between document comprehension, information retrieval, and cross-modal machine learning. The student is supposed to experiment with new approaches for captioning news images. We'll look into how the position of the images, or their relationship to other paragraphs, affects captioning performance. Can the model pay attention to the key details in the articles/images? Programming skill and good knowledge about deep learning is required.