Be Proud of Your Innovation

My final year project was about an image upscaling algorithm based on supervised machine learning. There were many cheap and small form factor camera modules available, however, these cameras can only output low resolution and frame rate video streams. After searching some scholarly articles, I found that passing the low-level information images to the last few layers can reconstruct a higher resolution image. I denote it as an information skipping structure. The algorithm can be used for image and video processing as it can outperform the traditional upscaling method in low-resolution images in a cost-effective way.

Mr. FUNG Siu Bong Benjamin
BSc (Hons) Engineering Physics
Faculty of Science

Domain Expertise:

Machine Learning and Artificial Intelligence

Benjamin's interest in machine learning and AI sparked during his second year of studying engineering physics. He took a machine learning course and explored the power of these technologies. His final year project focused on developing an image upscaling algorithm using machine learning methods.

Image Processing

Benjamin's final year project revolved around image upscaling, which falls under the domain of image processing. He aimed to enhance the quality of low-resolution footage from small form factor camera modules using machine learning algorithms.

High-Performance Computing and GPU Utilization

During his project, Benjamin faced technical challenges related to GPU memory, which led him to seek support from PolyU's High-Performance Computing Support Services. He utilized GPU resources, including a GTX 1080ti computer, to iterate his network faster for training the machine learning model.

Deep Learning

While working on his project, Benjamin learned about cutting-edge developments related to deep learning, including attention-based neural networks and generative adversarial networks. He utilized this knowledge to rebuild his network and improve the algorithm's performance.

Research and Literature Review

Benjamin emphasized the importance of literature review in research. He searched scholarly articles related to image upscaling and machine learning to gather insights and inspirations for his project. His ability to understand and apply research findings significantly contributed to his project's success.

Lifelong Learning Excellence:

Accountability and Motivation

Benjamin became a self-learner during his final year project. He had no prior experience working with image data, but he took the initiative to study deep learning concepts through videos and formal machine learning courses on platforms like YouTube and Coursera. This self-learning approach helped him understand the fundamentals and improve his project's algorithm..

Continuous Improvement and Learning from Mistakes

Benjamin encountered several difficulties during the development phase of his project, such as designing unsuccessful algorithms and facing technical challenges with GPU memory. He demonstrated adaptability by seeking solutions and adopting successful network designs from related articles. This experience taught him the importance of literature review and how it can provide valuable insights and directions for solving problems.
Benjamin's advice to incoming students is to try different things and explore various areas of interest. He himself was a physics student with an interest in computer science and board games. He explored multiple activities during his university life, including seeking research work, joining clubs, and pursuing internships. This curiosity and exploration helped him discover his passion for machine learning and AI.

Adaptability and Flexibility

Despite facing setbacks and disappointing results initially, Benjamin persisted and did not give up. He spent five months trying to make his original neural network design work. However, he later bounced back, learned from his mistakes, and restructured his approach to achieve better results. This showcases his resilience and determination.

Inspiring Quotes:

Explore More:

The pursuit of knowledge is a lifelong journey! To further expand your knowledge and continue your personal and professional growth. Click and explore the following learning resources:

Domain Knowledge OER