My name is Ali Hassan, and I am from Pakistan. My brief introduction is available here. I am working as Early Stage Researcher with double degree Doctoral Programme at Mid Sweden University (MIUN), Sweden, and Tampere University (TAU), Finland. My employer is MIUN and I started my PhD in October 1, 2021.
Research Focus
My research topic is “Dimension Reduction of Deep Network Models for Mobile Computational Plenoptics”. The aim of the research project is to compress and optimize deep learning architecture developed for Light Field applications, such that they can be deployed on mobile devices.
Research Progress
In the first phase, literature study were made on Deep Learning Compression and Optimization, along with study of the state-of-the-art solutions developed for Light Field application. Based on problem formulation, the popular state-of-the-art EPINET architecture developed for Light Field disparity estimation was compressed using convolution layer optimization technique, which result in significant complexity reduction by up to a factor of 3.89, while improving disparity estimation performance. The research work is published in IEEE Workshop on Multimedia Signal Processing, September 26-28, 2022.
In the second phase, theoretical and literature study is made on four other compression methods, including Neural Architecture Search, Pruning, Clustering, and Quantization. It is concluded from the implementation and experimental results of these compression techniques on image classification problem that the hybrid compression and collaborative optimization methods perform better than the individual compression method. In the future, the hybrid compression method and novel deep learning architectures will be proposed and the research finding will be submitted to a reputed international conferences or journals.
Academic Secondment
I have a six-month academic secondment at TAU, Finland. My secondment was started in January 2023. The research problems are formulated on Neural Architecture Search technique for Light Field application and implementation of the proposed solution is under-process. It is planned to implement and evaluate the proposed solution on different Light Field dataset during the secondment. During the secondment, the compulsory courses for PhD degree at TAU will be also studied.
Video
Publications
A. Hassan, M. Sjöström, T. Zhang and K. Egiazarian, “Light-Weight EPINET Architecture for Fast Light Field Disparity Estimation,” 2022 IEEE 24th International Workshop on Multimedia Signal Processing (MMSP), 2022, pp. 1-5, doi: 10.1109/MMSP55362.2022.9949378.