🌍 Spatial-SSRL: Spatial Reasoning with Vision-Language Models

Understanding 3D Spatial Relationships from 2D Images

✨ Upload an image and ask questions about spatial relationships, locations, and orientations! ✨

📖 Paper | 🏠 Github | 🤗 Spatial-SSRL-7B Model | 🤗 Spatial-SSRL-81k | 📰 Daily Paper

When enabled, the predefined format prompt is automatically concatenated to your question.

📸 Example Questions

Click on an example below to load it:
Complete Examples
Input Image Question Apply format prompt (default on)

About

This demo showcases spatial reasoning capabilities of vision-language models. The model can:

  • Understand 3D spatial relationships from 2D images
  • Reason about object locations (near/far, front/behind)
  • Analyze object orientations and facing directions
  • Provide step-by-step reasoning before answering

Citation

If you find this project useful, please kindly cite: