CLAUDEFORGE STUDIO

The vault — open access

The vault

Every document pulled in from the daily science alerts, newest first. Each has a short machine-written summary, a free copy anyone can open, and the date its link was last confirmed to work.

2 papers in neural radiance fields

neural radiance fieldsRead 2026-11-06

A Semi-Automatic Roof Type-Driven Workflow for LoD2 Building Reconstruction from Orthophotos and Photogrammetric Point Clouds

Nur Aini Lestari, Harintaka Harintaka, Ahmad Rasis Sholikin · 2026

This paper presents a semi-automatic workflow for building LoD2 3D city models, in which users digitize roof outlines from orthophotos and label roof types (flat, gable, hip, shed, pyramid), and rule-based code then reconstructs roof geometry using height and orientation derived from point cloud elevation layers. Applied to 371 buildings, the method outputs CityJSON models convertible to CityGML. On 30 buildings, roof-plane completeness averaged 100% and correctness and quality averaged 98.33%, with a geometric RMSE of 0.90 m.

Summary and sourcesFree copy via Alert linkLink checked 2026-11-06
neural radiance fieldsRead 2026-11-04

RADIANT: Object-Centric Feed-Forward Global Illumination

Sarah Zhang, Zhiyang Dou, Minghao Guo, Benjamin Tod Jones et al. · arXiv (Cornell University) · 2026

RADIANT is a feed-forward neural renderer that synthesizes global illumination by representing each object with a fixed budget of 32 surface-patch tokens, so attention cost depends on object count rather than mesh resolution. It uses an invariant form-factor attention to model light exchange between objects and a cross-attention view decoder to render images. On a 1.03M-scene benchmark, it reaches 32.2 dB PSNR at 600k training steps, 3.3 dB above RenderFormer.

Summary and sourcesFree copy via Alert linkLink checked 2026-11-04
← Back to the studio