Seeking open-source solutions for Image-to-3D reconstruction of entire rooms
Adventurous_Exit2832 · reddit · 2026-08-25
A user discovered GenRecon, which uses multi-view photos and camera poses combined with the TRELLIS.2 generative prior to reconstruct indoor spaces into textured PBR GLB meshes. They are asking for other open-source projects that can reconstruct entire rooms from photos, videos, or 360° panoramas, and are exploring pipelines using existing open-source models for depth estimation and scene reconstruction.
More from Multimodal
- MiniMax video generation: A sudden encounter with a dog on the way home — VasaFromParadise · 2026-08-25
- R2V workflow debate: Why do character sheets yield less consistency than single images? — SuspiciousRefuse8218 · 2026-08-25
- Canter 2B: Single-GPU trained model signals indie AI era — multimodalart · 2026-08-25
- AI filmmaking milestone: maintaining one visual world across a sequence — johnstro12 · 2026-08-25
- From storyboard to polished AI video: Liquid splash workflow — anthara_ai · 2026-08-25
- Hands-on: Pixel 11's Rambler is essentially Gemini inside Gboard — sqcai · 2026-08-25