This repository serves as a curated collection of research papers focused on 4D generation, a rapidly evolving field in computer graphics and AI. It targets researchers and practitioners interested in creating dynamic, three-dimensional scenes and objects from various inputs, offering a centralized resource for state-of-the-art techniques and advancements.

How It Works

The collection categorizes papers based on their core methodologies and applications within 4D generation. Key themes include controlling camera perspectives for video diffusion models, achieving multi-frame and multi-view consistency, leveraging distillation from existing video diffusion models, and enabling generation through reconstruction. Some papers also explore novel editing capabilities and physics-informed generation.

Quick Start & Requirements

This is a curated list of papers and does not involve direct code execution or installation. Links to official project pages, papers, and code repositories are provided for each entry.

Highlighted Details

• Comprehensive coverage of recent (2023-2024) advancements in 4D generation.
• Categorization by technical approach: Camera Control, Multi-view Consistency, Distillation, Reconstruction, Editing, and Physics.
• Includes papers utilizing Gaussian Splatting, diffusion models, and various forms of scene representation.
• Links to associated code, project pages, and videos for many entries.

Maintenance & Community

The repository is maintained by cwchenwang. As a list, it does not have an active community forum or roadmap in the traditional sense, but it reflects ongoing research trends.

Licensing & Compatibility

The repository itself is licensed under an unspecified license. Individual papers and their associated code repositories will have their own licenses, which must be consulted for usage and compatibility.

Limitations & Caveats

This repository is a passive collection of links and does not provide any executable code or integrated framework. Users must individually access and evaluate each linked paper and its associated resources.