Synthesis and generation for 3D architecture volume with generative modeling

International Journal of Architectural Computing, Journal Year: 2023, Volume and Issue: 21(2), P. 297 - 314

Published: April 3, 2023

Generative design in architecture has long been studied, yet most algorithms are parameter-based and require explicit rules, the solutions heavily experience-based. In absence of a real understanding generation process designing consensus evaluation matrices, empirical knowledge may be difficult to apply similar projects or deliver next generation. We propose workflow early phase synthesize generate building morphology with artificial neural networks. Using 3D models from financial district New York City as case study, this research shows that networks can capture implicit features styles input dataset create population coherent styles. constructed our database using two different data representation formats, voxel matrix signed distance function, investigate effect shape representations on performance shapes. A generative adversarial network an auto decoder were used volume. Our study establishes use learning inform solution. Results show both grasp forms them style data, between which function provides highest resolution results.

Language: Английский

---

A Study on Refraction Error Compensation Method for Underwater Spinning Laser Scanning Three-Dimensional Imaging

Sensors, Journal Year: 2024, Volume and Issue: 24(2), P. 343 - 343

Published: Jan. 6, 2024

Laser scanning 3D imaging technology, because it can obtain accurate three-dimensional surface data, has been widely used in the search for wrecks and rescue operations, underwater resource development, other fields. At present, conventional spinning laser system maintains a relatively fixed light window. However, low-light situations underwater, rotation of device causes some degree water fluctuation, which warps strip data that sensor receives about object’s surface. To solve this problem, research studies an makes use window (FWLS). A refraction error compensation algorithm is investigated based on fundamentals linear imaging, dynamic mathematical model established motion device. The results experiment analysis optimal environment indicate reconstructing radius decreased by 60% (from 2.5 mm to around 1 mm) when compensating measurement standard sphere with 20 mm. Moreover, compensated point cloud exhibit higher correspondence spherical cloud. Furthermore, we examine impact physical noise, distance, partial occlusion object inside authentic setting. This study good starting looking at refractive system. It also provides us ideas future methods.

Language: Английский

---

Vision-based recognition and geolocation of colored steel constructions in cities to promote urban management

Journal of Building Engineering, Journal Year: 2025, Volume and Issue: unknown, P. 111904 - 111904

Published: Jan. 1, 2025

Language: Английский

---

PAPRec: 3D Point Cloud Reconstruction Based on Prior-Guided Adaptive Probabilistic Network

Sensors, Journal Year: 2025, Volume and Issue: 25(5), P. 1354 - 1354

Published: Feb. 22, 2025

Inferring a complete 3D shape from single-view image is an ill-posed problem. The proposed methods often have problems such as insufficient feature expression, unstable training and limited constraints, resulting in low accuracy ambiguity reconstruction. To address these problems, we propose prior-guided adaptive probabilistic network for reconstruction, called PAPRec. In the stage, PAPRec encodes its corresponding prior into distribution point cloud distribution, respectively. then utilizes latent normalizing flow to fit two distributions obtains vector with rich cues. finally introduces consisting of diffusion model order decode cloud. Unlike methods, fully learns global local features objects by innovatively integrating guidance probability under optimization loss function combining prior, losses. experimental results on public ShapeNet dataset show that PAPRec, average, improves CD 2.62%, EMD 5.99% F1 4.41%, comparison several state-of-the-art methods.

Language: Английский

---

Modelling sunlight and shading distribution on 3D trees and buildings: Deep learning augmented geospatial data construction from street view images

Building and Environment, Journal Year: 2025, Volume and Issue: unknown, P. 112816 - 112816

Published: March 1, 2025

Language: Английский

---