Apple has released the 3D representation method Split-and-Scale Latent 3D Representations.
The story so far
Apple has released the Split-and-Scale Latent 3D Representations method, enhancing 3D scene detail capture by extending latent capabilities while keeping training costs manageable. The method represents 3D scenes as continuous functions, using a single pre-trained tokenizer to handle any sub-region, achieving higher detail reconstruction. Its inference-time extension method is comparable in quality to large models exceeding single GPU memory limits and supports high-quality generative modeling. This method significantly extends the capabilities of modern shape modeling, enabling compact encoding of 3D scenes and supporting powerful 3D generative models.
AI-written from the reports · Updated 22 hours ago
Coverage timeline
Follow the coverage to see every side of the story.
- Apple Machine Learning ResearchSplit-and-Scale Latent 3D Representations
Split-and-Scale enhances the ability to capture details in 3D scenes by expanding latent capabilities during inference while keeping training costs manageable. This method represents 3D scenes as continuous functions, using a single pre-trained tokenizer to process any sub-region of the scene, achieving higher-detail reconstruction. Its inference-time expansion method is comparable to large models exceeding single GPU memory limits in quality and supports high-quality generative modeling.
Heat over time
Not enough data yet to draw a trend.