At the recently concluded CES exhibition, Stability AI announced the launch of an innovative method called SPAR3D (Stable Point Aware3D). This two-stage 3D generation technology can create precise 3D objects from a single image in less than a second. The introduction of this technology provides game developers, product designers, and environment builders with a new way to prototype in 3D.
What makes SPAR3D unique is its ability to model both visible and hidden parts separately. When generating 3D objects, this method provides accurate geometric shapes and a complete 360-degree view, including details on the back of the objects. Users can also directly modify the generated 3D objects through real-time editing features, such as deleting, copying, stretching, and recoloring point clouds, greatly enhancing operational flexibility and creative space.
The workflow of SPAR3D can be divided into two stages. The first stage uses a lightweight point diffusion model to generate a sparse 3D point cloud, with extremely fast sample extraction; the second stage utilizes these point clouds and the input image to construct a high-detail mesh model. This design not only improves computational efficiency but also ensures output quality, overcoming the limitations of traditional single-image 3D reconstruction methods.
In performance testing, SPAR3D excelled on the GSO and Omniobject3D datasets, outperforming various existing state-of-the-art algorithms. The research team validated the effectiveness of this technology through multiple experiments, especially highlighting its advantages in handling image occlusion and detail reconstruction. SPAR3D uses explicit point clouds as an intermediate representation, allowing users to further edit 3D models, thus enhancing user experience.
Key features of SPAR3D include:
Real-time generation capability
Fast generation: Capable of generating a complete 3D object structure from a single image in less than a second.
Efficient editing: The speed of converting edited point clouds to final meshes is only 0.3 seconds.
Advanced editing features
Point cloud editing: Users can directly manipulate the point cloud, including deleting, copying, stretching, and adding features.
Recoloring: Provides the ability to adjust the colors of the 3D model, enhancing creative flexibility.
Accurate structure prediction
Complete structure prediction: Able to provide accurate geometric shapes and detailed 360-degree view predictions, including hidden areas.
High-resolution output: Generates high-resolution three-plane data through a three-plane transformer, capturing the texture and lighting of images.
Open usage policy
Wide application licensing: Supports both commercial and non-commercial use, allowing users to access and utilize the model across multiple platforms.
Developer-friendly: Weights can be downloaded via Hugging Face, code can be accessed on GitHub, or through API access.
Official introduction: https://stability.ai/news/stable-point-aware-3d
Key points:
🌟 SPAR3D can generate precise 3D objects from a single image in one second and supports real-time editing capabilities.
🔍 This technology employs a two-stage approach to generate point clouds and high-detail meshes, improving computational efficiency and output quality.
🚀 SPAR3D outperforms existing 3D reconstruction algorithms across multiple datasets, showcasing excellent performance and user editing capabilities.
