Thesis Format
Monograph
Degree
Master of Science
Program
Computer Science
Supervisor
Dr. Umair Rehman
Abstract
The rapid advancements in AI-generated models for image generation have transformed image creation from entertainment to e-commerce, making robust evaluation essential. We designed and statistically validated a human study for assessing the photorealism and image quality of AI-generated images. The study indicated that camera images show more realism, while AI images, such as those generated by DALL-E2, excel in quality. Further analysis showed that existing metrics deviate from human judgments. To address this, we developed the Global-Local Image Perceptual Score (GLIPS), which evaluates photorealistic quality using Vision Transformer-based attention for local similarity and Maximum Mean Discrepancy (MMD) for global distributional similarity. Our evaluation showed that GLIPS aligns more closely with human perception compared to traditional metrics like FID and SSIM for photorealism, although MS-SSIM outperformed in image quality. Additionally, we introduced the Interpolative Binning Scale (IBS) for human-assisted metric scaling with human scores.
Summary for Lay Audience
Artificial Intelligence (AI) technology has rapidly changed how images are created, and these images have been used in everything from movies and ads to online shopping. But with these advances, it's important to make sure that the images generated by AI look real and are of high quality. To do this, we conducted a study where people compared AI-generated images with photos taken by cameras. We found that while camera photos generally looked more realistic, some AI images, like those from DALL-E2, were rated higher in quality. However, when we compared these human opinions to existing computer-based metrics, we noticed that the metrics didn’t always match what people thought. To solve this, we designed a new way to measure the quality and realism of AI images, called the Global-Local Image Perceptual Score (GLIPS). This new method better aligns with human judgment, making it easier to assess how good AI-generated images really are. We also introduced a new scaling method to ensure that computer metrics more accurately reflect human opinions.
Recommended Citation
Aziz, Memoona Ms., "ASSESSMENT OF AI-GENERATED IMAGES USING COMPUTATIONAL METRICS AND HUMAN CENTRIC ANALYSIS" (2024). Electronic Thesis and Dissertation Repository. 10288.
https://ir.lib.uwo.ca/etd/10288
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.