Soheib works as Early Stage Researcher at Mid Sweden University (MIUN), Sweden, and his research topic is “Representation and Compression of Multi-sensor Video for Combined Human and Computer Vision Applications“.
You can read Soheib’s introduction here.
ESR12 Soheib Takhtardeshir
My name is Soheib Takhtardeshir. I am 30 years old. I come from Kurdistan-Iran. I started my bachelor’s in Electrical engineering in 2009 and, after finishing it in 2013, I did my military service for two years. After that, I joined FiberHome International Co. as a technical engineer in 2015. In 2019, I decided to continue my studies, and I was accepted as a master’s student in Digital electronic engineering. I finished my master’s in 2021 and started to find a suitable and exciting Ph.D. position.
Research Focus
My research focus is on compressing Light field Videos by deep learning-based methods for combined human and computer vision applications.
Progress
In the project’s first phase, a comprehensive analysis of existing literature was conducted to uncover the most advanced techniques in Light Field compression and to explore the use of Deep Learning in compression modeling. Moreover, based on identified deficiencies in current understanding, the research questions were crafted to investigate the potential of using deep learning for video compression as a means to compress light field images, creating pseudo video sequences. This approach aimed to increase familiarity with the field and serve as a basis for further exploration in the second phase of the research. With the objective in mind, a deep learning-based video compression approach is fine-tuned and adapted to effectively compress light field images, surpassing previous results in the field. The JPEG-Pleno dataset is utilized for both training and testing purposes.
Future work
The project’s second phase will build upon the insights and knowledge gained from the first phase. This phase will focus on utilizing deep learning end-to-end architectures, which are decoupled from the traditional hybrid coding paradigm and fully adopt signal-dependent compression to create a more robust representation of light field video. Additionally, the key applications of light field technology will be carefully evaluated, with a particular focus on understanding their unique features and characteristics.