scholarly journals 3DPlanNet: Generating 3D Models from 2D Floor Plan Images Using Ensemble Methods

Electronics ◽  
2021 ◽  
Vol 10 (22) ◽  
pp. 2729
Author(s):  
Sungsoo Park ◽  
Hyeoncheol Kim
Keyword(s):  
Ensemble Methods ◽  
Ground Truth ◽  
Structural Data ◽  
3D Models ◽  
Great Effort ◽  
Heuristic Methods ◽  
Floor Plan ◽  
Vector Data ◽  
Rule Based ◽  
Learning Data

Research on converting 2D raster drawings into 3D vector data has a long history in the field of pattern recognition. Prior to the achievement of machine learning, existing studies were based on heuristics and rules. In recent years, there have been several studies employing deep learning, but a great effort was required to secure a large amount of data for learning. In this study, to overcome these limitations, we used 3DPlanNet Ensemble methods incorporating rule-based heuristic methods to learn with only a small amount of data (30 floor plan images). Experimentally, this method produced a wall accuracy of more than 95% and an object accuracy similar to that of a previous study using a large amount of learning data. In addition, 2D drawings without dimension information were converted into ground truth sizes with an accuracy of 97% or more, and structural data in the form of 3D models in which layers were divided for each object, such as walls, doors, windows, and rooms, were created. Using the 3DPlanNet Ensemble proposed in this study, we generated 110,000 3D vector data with a wall accuracy of 95% or more from 2D raster drawings end to end.

scholarly journals Framework for Indoor Elements Classification via Inductive Learning on Floor Plan Graphs

2021 ◽  
Vol 10 (2) ◽  
pp. 97
Author(s):  
Jaeyoung Song ◽  
Kiyun Yu
Keyword(s):  
Neural Network ◽  
Inductive Learning ◽  
Network Data ◽  
Floor Plan ◽  
Vector Data ◽  
Adjacency Graph ◽  
Different Types ◽  
Image Pixels ◽  
Learning Frameworks ◽  
New Framework

This paper presents a new framework to classify floor plan elements and represent them in a vector format. Unlike existing approaches using image-based learning frameworks as the first step to segment the image pixels, we first convert the input floor plan image into vector data and utilize a graph neural network. Our framework consists of three steps. (1) image pre-processing and vectorization of the floor plan image; (2) region adjacency graph conversion; and (3) the graph neural network on converted floor plan graphs. Our approach is able to capture different types of indoor elements including basic elements, such as walls, doors, and symbols, as well as spatial elements, such as rooms and corridors. In addition, the proposed method can also detect element shapes. Experimental results show that our framework can classify indoor elements with an F1 score of 95%, with scale and rotation invariance. Furthermore, we propose a new graph neural network model that takes the distance between nodes into account, which is a valuable feature of spatial network data.

scholarly journals RobotP: A Benchmark Dataset for 6D Object Pose Estimation

Sensors ◽  
2021 ◽  
Vol 21 (4) ◽  
pp. 1299
Author(s):  
Honglin Yuan ◽  
Tim Hoogenkamp ◽  
Remco C. Veltkamp
Keyword(s):  
Pose Estimation ◽  
Ground Truth ◽  
3D Models ◽  
Depth Image ◽  
Great Success ◽  
Estimation Algorithms ◽  
Depth Images ◽  
Object Pose Estimation ◽  
Image Pairs ◽  
Bounding Boxes

Deep learning has achieved great success on robotic vision tasks. However, when compared with other vision-based tasks, it is difficult to collect a representative and sufficiently large training set for six-dimensional (6D) object pose estimation, due to the inherent difficulty of data collection. In this paper, we propose the RobotP dataset consisting of commonly used objects for benchmarking in 6D object pose estimation. To create the dataset, we apply a 3D reconstruction pipeline to produce high-quality depth images, ground truth poses, and 3D models for well-selected objects. Subsequently, based on the generated data, we produce object segmentation masks and two-dimensional (2D) bounding boxes automatically. To further enrich the data, we synthesize a large number of photo-realistic color-and-depth image pairs with ground truth 6D poses. Our dataset is freely distributed to research groups by the Shape Retrieval Challenge benchmark on 6D pose estimation. Based on our benchmark, different learning-based approaches are trained and tested by the unified dataset. The evaluation results indicate that there is considerable room for improvement in 6D object pose estimation, particularly for objects with dark colors, and photo-realistic images are helpful in increasing the performance of pose estimation algorithms.

scholarly journals Monitoring and Computation of the Volumes of Stockpiles of Bulk Material by Means of UAV Photogrammetric Surveying

2019 ◽  
Vol 11 (12) ◽  
pp. 1471 ◽  
Author(s):  
Grazia Tucci ◽  
Antonio Gebbia ◽  
Alessandro Conti ◽  
Lidia Fiorini ◽  
Claudio Lubello
Keyword(s):  
Bulk Material ◽  
Laser Scanner ◽  
Ground Truth ◽  
Point Clouds ◽  
3D Models ◽  
Satellite System ◽  
3D Data ◽  
Current Survey ◽  
Close Range ◽  
Global Navigation Satellite

The monitoring and metric assessment of piles of natural or man-made materials plays a fundamental role in the production and management processes of multiple activities. Over time, the monitoring techniques have undergone an evolution linked to the progress of measure and data processing techniques; starting from classic topography to global navigation satellite system (GNSS) technologies up to the current survey systems like laser scanner and close-range photogrammetry. Last-generation 3D data management software allow for the processing of increasingly truer high-resolution 3D models. This study shows the results of a test for the monitoring and computing of stockpile volumes of material coming from the differentiated waste collection inserted in the recycling chain, performed by means of an unmanned aerial vehicle (UAV) photogrammetric survey and the generation of 3D models starting from point clouds. The test was carried out with two UAV flight sessions, with vertical and oblique camera configurations, and using a terrestrial laser scanner for measuring the ground control points and as ground truth for testing the two survey configurations. The computations of the volumes were carried out using two software and comparisons were made both with reference to the different survey configurations and to the computation software.

scholarly journals A New Approach to 3D Modeling of Inhomogeneous Populations of Viral Regulatory RNA

Viruses ◽  
2020 ◽  
Vol 12 (10) ◽  
pp. 1108
Author(s):  
Patrick S. Osmer ◽  
Gatikrushna Singh ◽  
Kathleen Boris-Lawrie
Keyword(s):  
Tertiary Structure ◽  
Ensemble Methods ◽  
3D Models ◽  
Primer Binding Site ◽  
Regulatory Rna ◽  
Input Constraints ◽  
Rna Motifs ◽  
Genetic Studies ◽  
New Approach ◽  
Physical Context

Tertiary structure (3D) is the physical context of RNA regulatory activity. Retroviruses are RNA viruses that replicate through the proviral DNA intermediate transcribed by hosts. Proviral transcripts form inhomogeneous populations due to variable structural ensembles of overlapping regulatory RNA motifs in the 5′-untranslated region (UTR), which drive RNAs to be spliced or translated, and/or dimerized and packaged into virions. Genetic studies and structural techniques have provided fundamental input constraints to begin predicting HIV 3D conformations in silico. Using SimRNA and sets of experimentally-determined input constraints of HIVNL4-3 trans-activation responsive sequence (TAR) and pairings of unique-5′ (U5) with dimerization (DIS) or AUG motifs, we calculated a series of 3D models that differ in proximity of 5′-Cap and the junction of TAR and PolyA helices; configuration of primer binding site (PBS)-segment; and two host cofactors binding sites. Input constraints on U5-AUG pairings were most compatible with intramolecular folding of 5′-UTR motifs in energetic minima. Introducing theoretical constraints predicted metastable PolyA region drives orientation of 5′-Cap with TAR, U5 and PBS-segment helices. SimRNA and the workflow developed herein provides viable options to predict 3D conformations of inhomogeneous populations of large RNAs that have been intractable to conventional ensemble methods.

Holocene dyke-induced surface deformation at Krafla (Iceland) revealed by UAV-based high resolution 3D models

2020 ◽  
Author(s):  
Alessandro Tibaldi ◽  
Elena Russo ◽  
Luca Fallati
Keyword(s):  
Field Survey ◽  
Surface Deformation ◽  
Structural Data ◽  
3D Models ◽  
Surface Model ◽  
Normal Faulting ◽  
Volcanic Zone ◽  
The North ◽  
Very High ◽  
Main Deformation

<p>We analysed at very high detail the surface deformation along a volcanotectonic structure in the Krafla Fissure Swarm, located in the North Iceland Rift. The structure affects the Pleistocene Hituholar volcano and 12 ka old lava flows. The work has been carried out through the Structure from Motion technique (SfM) applied to UAV surveys, integrated with a lithostratigraphic and structural field survey. The resulting Orthomosaic and Digital Surface Model (DSM) have a resolution of 2.6 and 10 cm, respectively. The zone of deformation is characterised by topographic bulging, parallel extension fractures, and narrow grabens with locally floor uplift, which can be explained as the effect of shallow propagation of a dyke northward from the Krafla magma chamber. In fact, the study area has been interested by northward dyke propagation from the central Krafla volcano during several rifting events, among which the recentmost occurred in 1975-1984 (Krafla fire). The analysis of the very wide area covered by our UAV surveys indicates that changes in the pattern of surface deformation occur in correspondence of contacts between deposits with different rheological properties: the transition from very stiff lavas to soft hyaloclastites produces a change from extension fracturing to normal faulting. Moreover, we detected a series of extension fractures with NE-SW strike and left-lateral slip component, and NNW-SSE strike and right-lateral component, which are rotated clockwise and anticlockwise respect to the main NNE-SSW graben trend, and extend outward to the sides of the main deformation zone up to 17 m. We interpret these structures as originated in front of the dyke tip during its propagation and being successively bypassed by the dyke advancement. In case of an active volcanic zone, the comprehension of the surface deformation and of the significance of strike-slip faulting occurrence can help to determine how and where magma is propagating. Thus, these evidences may help to decipher geophysical data and surface structural data during volcano monitoring.</p>

scholarly journals A Synthetic LiDAR Scanner for VTK

2009 ◽  
Author(s):  
David Doria
Keyword(s):  
3D Model ◽  
Real Life ◽  
Ground Truth ◽  
Laser Pulses ◽  
3D Models ◽  
Light Detection And Ranging ◽  
Data Sets ◽  
Model Surface ◽  
Point Data ◽  
Reflecting Surface

In recent years, Light Detection and Ranging (LiDAR) scanners have become more prevalent in the scientific community. They capture a “2.5-D” image of a scene by sending out thousands of laser pulses and using time-of-flight calculations to determine the distance to the first reflecting surface in the scene. Rather than setting up a collection of objects in real life and actually sending lasers into the scene, one can simply create a scene out of 3d models and “scan” it by casting rays at the models. This is a great resource for any researchers who work with 3D model/surface/point data and LiDAR data. The synthetic scanner can be used to produce data sets for which a ground truth is known in order to ensure algorithms are behaving properly before moving to “real” LiDAR scans. Also, noise can be added to the points to attempt to simulate a real LiDAR scan for researchers who do not have access to the very expensive equipment required to obtain real scans.

MEASURING SIMILARITY BETWEEN BIOMEDICAL DATA BY USING FURIA ENSEMBLES RULE-BASED CLASSIFICATION

2020 ◽  
Vol 2 (1) ◽  
pp. 116-127
Author(s):  
Simon Fong
Keyword(s):  
Similarity Measures ◽  
Fuzzy Rule ◽  
Ensemble Methods ◽  
Biomedical Data ◽  
Rule Based ◽  
Reasoning System ◽  
Evaluation Approach ◽  
Common Task ◽  
Test Instance ◽  
The Difference

Similarity measures are essential to solve many pattern recognition problems such as classification, clustering, and retrieval problems. Various distance/similarity measures that is applicable to compare two probability density functions. Data comparison is widely used field in our society nowadays, and it is a very import part. To compare two objects is a common task that people from all walks of life would do. People always want or need to find the similarity between two different objects or the difference between two similar objects. Some different data may share some similarity in some given attribute(s). To compare with two datasets based on attributes by classification algorithms, for the attributes, we need to select them out by rules and the system is known as rule-based reasoning system or expert system which classifies a given test instance into a particular outcome from the learned rules. The test instance carries multiple attributes, which are usually the values of diagnostic tests. In this article, we are proposing a classifier ensemble-based method for comparison of two datasets or one dataset with different features. The ensemble data mining learning methods are applied for rule generation, and a multi-criterion evaluation approach is used for selecting reliable rules over the results of the ensemble methods. The efficacy of the proposed methodology is illustrated via an example of two disease datasets; it is a combined dataset with the same instances and normal attributes but the class in strictly speaking. This article introduces a fuzzy rule-based classification method called FURIA, to get the relationship between two datasets by FURIA rules. And find the similarity between these two datasets.

scholarly journals 3D printing of crystallographic models and open access databases

2014 ◽  
Vol 70 (a1) ◽  
pp. C1278-C1278 ◽  
Author(s):  
Werner Kaminsky ◽  
Trevor Snyder ◽  
Peter Moeck
Keyword(s):  
Open Access ◽  
3D Printing ◽  
College Education ◽  
Structural Data ◽  
Data Bank ◽  
3D Models ◽  
Crystal Defects ◽  
Crystallographic Structure ◽  
Performance Improvements ◽  
Basic Infrastructure

Although introduced 30 years ago, cost and performance improvements have only recently made 3D printing affordable. The industry wide input file format for 3D printers incorporates explicit mesh - `STL' data. Molecules and crystal structures, when including symmetry, crystal morphologies, or crystal defects are encoded in the parametrical `CIF' syntax. Free software for converting directly CIF data to STL files has just been developed, available online [1]. First examples of printed 3D models from STL-files created with these programs include molecules of sucrose, herapathite [2a], caffeine, humulone [2b], an alpha-quartz crystal and its Japanese {112} twin or a brilliant cut diamond. Far more CIF encoded models are available, even open access. The Crystallography Open Database (COD) features over 245,000 entries and has recently developed into the world's premier open-access source for structures of small to medium unit cell-sized inorganic and molecular crystals [3a], complementing the well-established open-access Worldwide Protein Data Bank [3b]. The Cambridge Crystallographic Data Centre in the United Kingdom provides crystal structure data of small (organic) molecules free for bona fide research [3c]. Structural data on inorganic crystals, metals and alloys can be obtained free of charge from the Inorganic Material Database (AtomWork) [3d]. Related to the COD, the crystallographic open-access databases [3e] ("COD offspring") provide CIF data for interdisciplinary college education. With this basic infrastructure in place, any interested college educator may print out her or his favorite crystallographic structure model in 3D and use it in hands on class room demonstrations [3f].

scholarly journals Real-time surface manipulation with $$C^{1}$$ continuity through simple and efficient physics-based deformations

2021 ◽  
Author(s):  
Shuangbu Wang ◽  
Nan Xiang ◽  
Yu Xia ◽  
Lihua You ◽  
Jianjun Zhang
Keyword(s):  
Unit Circle ◽  
Real Time ◽  
Ground Truth ◽  
3D Models ◽  
Surface Shape ◽  
Thin Plates ◽  
Coordinate Space ◽  
Order Partial Differential Equation ◽  
Deformation Region ◽  
Parametric Plane

AbstractWe present a novel but simple physics-based method to interactively manipulate surface shapes of 3D models with $$ C^1 $$ C 1 continuity in real time. A fourth-order partial differential equation involving a sculpting force originating from elastic bending of thin plates is proposed to define physics-based deformations and achieve $$ C^1 $$ C 1 continuity at the boundary of deformation regions. In order to obtain real-time physics-based surface manipulation, we construct a mapping relationship between a deformation region in a 3D coordinate space and a unit circle on a 2D parametric plane, formulate corresponding $$ C^1 $$ C 1 continuous boundary conditions for the unit circle, and obtain a simple analytical solution to describe the physics-based deformation in the unit circle caused by a sculpting force. After that, the obtained physics-based deformation is mapped back to the 3D coordinate space, and added to the original surface to create a new surface shape with $$ C^1 $$ C 1 continuity at the boundary of the deformation region. We also develop an interactive user interface as a plug-in of the 3D modelling software package Maya to achieve real-time surface manipulation. The effectiveness, easiness, real-time performance, and better realism of our proposed method is demonstrated by testing surface deformations on several 3D models and comparing with other methods and ground-truth deformations.

Sign in / Sign up

Export Citation Format

Share Document