Project Abstract

Image enhancement is important among computer vision operations and it aims to improve the resolution and quality of low-resolution images. The existing methods often suffer from reduction of details and added artifacts whereas the AI-enabled approaches are costly computationally and lack standardized performance evaluation metrics. Due to the problems the project is made to resolve, it proposes a supportive AI device of image super-sampling able to build high-resolution images that are lifelike and still retain characteristic features. The project works with KDEF as a database for training and testing, while to standing on strong deep learning structures and optimization algorithms. This project focuses on the research questions in the area of deep learning model that are aimed at finding the effective architecture for bringing depth and better quality in images, reducing the number of parameters and time of run, and evaluating the level of performance gain achieved. Data evaluation indicates that deep learning algorithms play a key role in the image super-resolution area, where the methodology has been usually proposed for providing both recovery and increasing the quality of the image. Then a CNN model with Adam optimizer and MSE loss function was designed and retrained. Intense manpower is focused on optimization of attention rate and speed of learning on large images. Metrics like SSIM index which are designed to evaluate the model’s ability to produce realistic images or high-quality ones are used. The performance is beyond expectation: the AI does the thing which it was designed to do- -restoring the image with a very high resolution. Auxiliary indicators and performance charts display the efficiency of the proposed algorithm in increasing image sharpness while precluding any second-rate visual elements. As a result of this project, it has made one step towards image super-resolution technology development as a system that can be used in many of the situations that take place in the real world.

Keywords: AI-enabled Approach, Deep Learning, AI-enabled Approach

 Conference Details

Session: Presentation Stream 17 at Presentation Slot 10

Location: GH011 at Wednesday 8th 09:00 – 12:30

Markers: Arnold Beckmann, Alex Warren (GTA)

Course: MSc Advanced Computer Science, Masters PG

Future Plans: I’m looking for work