yslan

/

STream3R

like 4

This PR adds a comprehensive model card for STream3R, a novel approach to scalable sequential 3D reconstruction with Causal Transformer.

It includes:

• Relevant metadata such as pipeline_tag: image-to-3d, library_name: stream3r, and license: other.
• The paper title, abstract, and links to the official Hugging Face paper and project page.
• A link to the GitHub repository for the code.
• Detailed usage examples for inference, including streaming inference with KV cache.
• Information regarding the Gradio demo.
• Quantitative results and GPU memory/runtime statistics from the paper.
• Citation information for the paper.
• Details about the custom license.

This update will significantly improve the discoverability, usability, and documentation of the model on the Hugging Face Hub.