visual consistency
Visual consistency ensures that generated or synthesized images maintain a coherent style and adhere to expected visual features, which is essential in applications like content generation and image transformation.
- Enhancing Diffusion-based Unrestricted Adversarial Attacks via Adversary Preferences Alignment
- Fine-Grained Preference Optimization Improves Spatial Reasoning in VLMs
- FlexWorld: Progressively Expanding 3D Scenes for Flexible-View Exploration
- ImageSentinel: Protecting Visual Datasets from Unauthorized Retrieval-Augmented Image Generation
- RoomEditor: High-Fidelity Furniture Synthesis with Parameter-Sharing U-Net
- VideoRFT: Incentivizing Video Reasoning Capability in MLLMs via Reinforced Fine-Tuning