Perception-Guided AI for reliable image generation

Overview

Generative Artificial Intelligence (AI) can now create highly realistic images from text descriptions, but it still struggles with complex scenes involving multiple objects, people, and interactions, often producing unrealistic or inconsistent results. Improving the reliability of AI-generated images is essential for their safe adoption across creative industries, education, healthcare, and digital media.

This PhD project will develop novel AI methods to improve the reliability of AI-generated images in complex scenes. The research will first establish a benchmark dataset containing human annotations of image defects and then investigate how AI can learn from human perception to identify errors, localise defects, and guide the generation of more realistic and trustworthy images.

The successful candidate will join the Computer Vision and Image Processing Group at the University of Dundee, working on cutting-edge research in visual quality assessment and generative AI. The outcomes of this project will contribute to the development of more reliable, trustworthy, and human-centred generative AI technologies for real-world applications.

Diversity statement

Our research community thrives on the diversity of students and staff which helps to make the University of Dundee a UK university of choice for postgraduate research. We welcome applications from all talented individuals and are committed to widening access to those who have the ability and potential to benefit from higher education.