Nebius Token Factory
Mini toolsVIGA
By Fugtemypt123
VIGA (Vision-as-Inverse-Graphics Agent) is a multimodal agent that treats vision as inverse graphics by rebuilding an input image as a 3D program in Blender. It alternates generator and verifier roles in an analysis-by-synthesis loop, using interleaved language, vision and memory to infer objects, materials, lighting, physics and interactions, and shows strong generalization on BlenderGym and other 3D reasoning benchmarks.
New
Image
Gen 4
Overview
VIGA is a vision-as-inverse-graphics agent that rebuilds a single image as an editable 3D Blender scene, alternating generator and verifier roles with interleaved multimodal reasoning to capture objects, layout, physics and interactions.
About Fugtemypt123
VIGA creators :
Shaofeng Yin
Jiaxin Ge
Zora Zhiruo Wang
Xiuyu Li
Michael J. Black
Trevor Darrell
Angjoo Kanazawa1
Haiwen Feng
VIGA is an analysis-by-synthesis code agent for programmatic visual reconstruction. It approaches vision-as-inverse-graphics through an iterative loop of generating, rendering, and verifying scenes against target images.
Tools using VIGA
No tools found for this model yet.
Last updated: January 27, 2026
