Visual Appearance

Visual appearance research focuses on understanding and manipulating how images and videos are perceived, aiming to improve image editing, quality assessment, and object recognition. Current research emphasizes disentangling content from appearance using diffusion models and other deep learning architectures, enabling sophisticated edits to real images and videos while preserving temporal consistency and mitigating biases. These advancements have significant implications for various applications, including robotics, virtual try-ons, and improving the fairness and accuracy of vision-language models.