Def

Abstract

We propose Deep Estimators of Features (DEFs), a learning-based framework for predicting sharp geometric features in sampled 3D shapes. Differently from existing data-driven methods, which reduce this problem to feature classification, we propose to regress a scalar field representing the distance from point samples to the closest feature line on local patches. Our approach is the first that scales to massive point clouds by fusing distance-to-feature estimates obtained on individual patches. We extensively evaluate our approach against related state-of-the-art methods on newly proposed synthetic and real-world 3D CAD model benchmarks. Our approach not only outperforms these (with improvements in Recall and False Positives Rates), but generalizes to real-world scans after training our model on synthetic data and fine-tuning it on a small dataset of scanned data. We demonstrate a downstream application, where we reconstruct an explicit representation of straight and curved sharp feature lines from range scan data. We make code, pre-trained models, and our training and evaluation datasets available at https://github.com/artonson/def.

Efficient building facade structure extraction method using image‐based laser point cloud

“…Recently, machine and deep learning methods, such as hierarchical clustering, random forests, and adversarial networks, are introduced into the building facade extraction from 3D PCD directly (Teruggi et al, 2020; Wang, Chan, et al, 2020; Zhang et al, 2020). For the machine and deep learning‐based direct extraction methods, the structure descriptor, such as geometric features of building facades, is built relying on the rigid training data, which is used to classify facade structures from 3D PCD (Matveev et al, 2022; Mirzaei et al, 2022). However, the extraction results mirror the training data, which is generally subjective and huge, reducing the robustness and practicality of these methods (Wang et al, 2022).…”