A Dual-UNet Diffusion Framework for Personalized Panoramic Generation.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Computer Graphics

Background:

• Current single-image generation methods lack support for coherent 360° panoramas.
• Existing panorama generation models do not offer customization capabilities.
• Challenges include weakly correlated reference images and minor object presence in panoramic scenes.

Purpose of the Study:

• To propose a diffusion-based framework for customized multi-view image generation.
• To address limitations in generating coherent and customized panoramic scenes for immersive applications.

Main Methods:

• A decoupled feature injection mechanism within a dual-UNet architecture.
• Hybrid attention mechanism for deep fusion of reference features and multi-view representations.
• Data augmentation for viewpoint-adaptive pose adjustments and panoramic coordinates for multi-view attention.

Main Results:

• The proposed framework effectively handles weakly correlated reference images.
• Demonstrates successful integration of spatial information for denoising.
• Generates coherent and high-quality customized multi-view images.

Conclusions:

• The diffusion-based framework advances customized panoramic image generation.
• Enables coherent multi-view image synthesis for immersive applications.
• Offers a robust solution for challenges in panoramic scene generation.