Realsee Argus: LiDAR-Free Panoramic 3D Reconstruction via Smartphone
机器之心 · wechat · 2026-07-06
Argus, developed by the Realsee team, has been accepted to ECCV 2026. Designed for indoor panoramic images, it directly predicts camera pose, metric depth, and point cloud reconstruction from sparse, unordered panoramic photos, providing more stable and accurate geometric constraints for 3D Gaussian Splatting (3DGS). This means production-grade 3DGS reconstruction may no longer require LiDAR; capturing with just a smartphone or panoramic camera makes the process lighter and more cost-effective.
On the Realsee3D benchmark, Argus achieves SOTA in pose and geometric accuracy: compared to MapAnything360, it reduces global pose error by about 28% on the real subset and 69% on the synthetic subset. After training on tens of millions of data points for common residential indoor scenes, the error drops to as low as 2.5cm, approaching LiDAR's 2cm. Furthermore, it avoids common LiDAR issues like multi-echo trailing and inaccurate ranging for glass, mirrors, and black objects.
More from Multimodal
- FLUX.2 Klein Drifts Hard on Character Expressions While Free Gemini Holds Likeness — wacomlover · 2026-09-11
- Tencent Hunyuan releases AuK code and weights on GitHub with ComfyUI and fine-tuning support — aigclink · 2026-09-11
- Creator turns Bahamut vs Tiamat rivalry into an AI cinematic battle with Midjourney, GPT Image 2 and Seedance — azed_ai · 2026-09-11
- invideo launches AI agent-powered editor to automate repetitive editing tasks — azed_ai · 2026-09-11
- fable 5.1 recreates The Starry Night with 256,157 JavaScript brush strokes — cedric_chee · 2026-09-11
- GPT-6 Astra + Hyper3D Rodin MCP Generates 3D Assets in One Agent Flow — ahuja_priyank · 2026-09-11