Project page for "Planning with the Views" (arXiv:2605.29563): ViewAgent, view planning with multi-turn VLM agents, and ViewSuite, a 6-DoF view planning benchmark on real ScanNet scenes. View planning probes active perception with vision-language foundation models: an agent for embodied spatial reasoning chooses where to look next, separately from where to go.
- Live site: https://viewagent.github.io/
- Paper: https://arxiv.org/abs/2605.29563
- Code: https://github.com/mll-lab-nu/ViewAgent
- Dataset: https://huggingface.co/datasets/MLL-Lab/viewsuite
- Models: https://huggingface.co/collections/MLL-Lab/viewagent-models
- Lab: MLL Lab, Northwestern University (PI: Manling Li)
Built with Astro; deployed to GitHub Pages via the workflow in .github/workflows/deploy.yml.