Data-Efficient Machine Learning on 3D Engineering Data

2021 ◽  
pp. 1-14
Author(s):  
Vencia D Herzog ◽  
Stefan Suwelack
Keyword(s):  
Learning Strategies ◽  
Deep Neural Networks ◽  
Small Data ◽  
Design Synthesis ◽  
3D Shape ◽  
Engineering Data ◽  
3D Data ◽  
Efficient Machine ◽  
Data Efficiency ◽  
Shape Data

Abstract Decisions in engineering design are closely tied to the 3D shape of the product. Limited availability of 3D shape data and expensive annotation present key challenges for using Artificial Intelligence in product design and development. In this work we explore transfer learning strategies to improve the data-efficiency of geometric reasoning models based on deep neural networks as used for tasks such as shape retrieval and design synthesis. We address the utilization of problem- related and un-annotated 3D data to compensate for small data volumes. Our experiments show promising results for knowledge transfer on mechanical component benchmarks.

scholarly journals CitySAC: A Query-Able CityGML Compression System

Smart Cities ◽  
2019 ◽  
Vol 2 (1) ◽  
pp. 106-117
Author(s):  
Chengxi Siew ◽  
Pankaj Kumar
Keyword(s):  
Data Storage ◽  
Spatial Data ◽  
City Planning ◽  
Data Transfer ◽  
Original Data ◽  
File Size ◽  
Compression Technique ◽  
3D Data ◽  
Data Infrastructures ◽  
Data Efficiency

Spatial Data Infrastructures (SDIs) are frequently used to exchange 2D & 3D data, in areas such as city planning, disaster management, urban navigation and many more. City Geography Mark-up Language (CityGML), an Open Geospatial Consortium (OGC) standard has been developed for the storage and exchange of 3D city models. Due to its encoding in XML based format, the data transfer efficiency is reduced which leads to data storage issues. The use of CityGML for analysis purposes is limited due to its inefficiency in terms of file size and bandwidth consumption. This paper introduces XML based compression technique and elaborates how data efficiency can be achieved with the use of schema-aware encoder. We particularly present CityGML Schema Aware Compressor (CitySAC), which is a compression approach for CityGML data transaction within SDI framework. Our test results show that the encoding system produces smaller file size in comparison with existing state-of-the-art compression methods. The encoding process significantly reduces the file size up to 7–10% of the original data.

An evaluation methodology for 3D deep neural networks using visualization in 3D data classification

2019 ◽  
Vol 33 (3) ◽  
pp. 1333-1339
Author(s):  
Hyun-Tae Hwang ◽  
Soo-Hong Lee ◽  
Hyung Gun Chi ◽  
Nam Kyu Kang ◽  
Hyeon Bae Kong ◽  
...  
Keyword(s):  
Neural Networks ◽  
Deep Neural Networks ◽  
Data Classification ◽  
Evaluation Methodology ◽  
3D Data

scholarly journals 3D DIGITIZING TECHNOLOGY IN PRODUCT REVERSE DESIGN

2011 ◽  
Author(s):  
Qingjin Peng ◽  
Hector Sanchez
Keyword(s):  
Shape Recovery ◽  
Three Dimensional ◽  
Laser Scanner ◽  
Coordinate Measuring Machine ◽  
Cost Effective ◽  
3D Shape ◽  
3D Data ◽  
Cost Effective Approach ◽  
Measuring Machine ◽  
And Performance

The reverse design develops new products based on the improvement of existing products. The shape recovery of three-dimensional (3D) objects is the basis of the product reverse design. 3D digitization technology is an important tool for the 3D shape recovery. This paper analyses the current 3D data acquisition technology. The accuracy and performance of the 3D laser scanner is evaluated. A cost-effective approach is proposed to recover 3D shape of objects using a structured-light technique. Details of the proposed method are described. Application examples are presented. The accuracy is evaluated using a coordinate measuring machine.

scholarly journals Registration and Deformation of 3D Shape Data through Parameterized Formulation

2007 ◽  
Vol 3 ◽  
pp. 404-423
Author(s):  
Tomohito Masuda ◽  
Katsushi Ikeuchi
Keyword(s):  
3D Shape ◽  
Shape Data

scholarly journals 4D Atlas: Statistical Analysis of the Spatiotemporal Variability in Longitudinal 3D Shape Data

2021 ◽  
Author(s):  
HAMID LAGA ◽  
Marcel Padilla ◽  
Ian H. Jermyn ◽  
Sebastian Kurtek ◽  
Mohammed Bennamoun ◽  
...  
Keyword(s):  
Statistical Analysis ◽  
Building Blocks ◽  
Shape Space ◽  
Spatiotemporal Variability ◽  
3D Shape ◽  
Execution Rates ◽  
And Performance ◽  
Body Shapes ◽  
Spatiotemporal Registration ◽  
Shape Data

We propose a novel framework to learn the spatiotemporal variability in longitudinal 3D shape data sets, which contain observations of subjects that evolve and deform over time. This problem is challenging since surfaces come with arbitrary parameterizations and thus, they need to be spatially registered onto each others. Also, different deforming subjects, hereinafter referred to as 4D surfaces, evolve at different speeds and thus, they need to be temporally aligned onto each others. We solve this spatiotemporal registration problem using a Riemannian approach. We treat a 3D surface as a point in a shape space equipped with an elastic Riemmanian metric that measures the amount of bending and stretching that the surfaces undergo. A 4D surface can then be seen as a trajectory in this space. With this formulation, the statistical analysis of 4D surfaces can be cast as the problem of analyzing trajectories, or 1D curves, embedded in a nonlinear Riemannian manifold. However, performing the spatiotemporal registration, and subsequently computing statistics, on such nonlinear spaces is not straightforward as they rely on complex nonlinear optimizations. Our core contribution is the mapping of the surfaces to the space of Square-Root Normal Fields (SRNF) where the L2 metric is equivalent to the partial elastic metric in the space of surfaces. Thus, by solving the spatial registration in the SRNF space, the problem of analyzing 4D surfaces becomes the problem of analyzing trajectories embedded in the SRNF space, which has a Euclidean structure. In this paper, we develop the building blocks that enable such analysis. These include: (1) the spatiotemporal registration of arbitrarily parameterized 4D surfaces even in the presence of large elastic deformations and large variations in their execution rates, (2) the computation of geodesics between 4D surfaces, (3) the computation of statistical summaries, such as means and modes of variation, of collections of 4D surfaces, and (4) the synthesis of random 4D surfaces. We demonstrate the utility and performance of the proposed framework using 4D facial surfaces and 4D human body shapes.

scholarly journals Learning with small amount of data on various task

2020 ◽  
Author(s):  
Stefanie
Keyword(s):  
Deep Neural Networks ◽  
Learning Strategy ◽  
Teaching Strategy ◽  
Training Sample ◽  
Learning To Teach ◽  
Training Data ◽  
Training Dataset ◽  
Small Data ◽  
Training Samples ◽  
Teacher Model

As a student, I am learning knowledge with the help of teachers and the teacher plays a crucial role in our life. A wonderful instructor is able to teach a student with appropriate teaching materials. Therefore, in this project, I explore a teaching strategy called learning to teach (L2T) in which a teacher model could provide high-quality training samples to a student model. However, one major problem of L2T is that the teacher model will only select a subset of the training dataset as the final training data for the student. Learning to teach small-data learning strategy (L2TSDL) is proposed to solve this problem. In this strategy, the teacher model will calculate the importance score for every training sample and help students to make use of all training samples. To demonstrate the advantage of the proposed approach over L2T, I take the training of different deep neural networks (DNN) on image classification task as an exampleand show that L2TSDL could achieve good performance on both large and small dataset.

scholarly journals 4D Atlas: Statistical Analysis of the Spatiotemporal Variability in Longitudinal 3D Shape Data

2021 ◽  
Author(s):  
HAMID LAGA ◽  
Marcel Padilla ◽  
Ian H. Jermyn ◽  
Sebastian Kurtek ◽  
Mohammed Bennamoun ◽  
...  
Keyword(s):  
Statistical Analysis ◽  
Building Blocks ◽  
Shape Space ◽  
Spatiotemporal Variability ◽  
3D Shape ◽  
Execution Rates ◽  
And Performance ◽  
Body Shapes ◽  
Spatiotemporal Registration ◽  
Shape Data

We propose a novel framework to learn the spatiotemporal variability in longitudinal 3D shape data sets, which contain observations of subjects that evolve and deform over time. This problem is challenging since surfaces come with arbitrary parameterizations and thus, they need to be spatially registered onto each others. Also, different deforming subjects, hereinafter referred to as 4D surfaces, evolve at different speeds and thus, they need to be temporally aligned onto each others. We solve this spatiotemporal registration problem using a Riemannian approach. We treat a 3D surface as a point in a shape space equipped with an elastic Riemmanian metric that measures the amount of bending and stretching that the surfaces undergo. A 4D surface can then be seen as a trajectory in this space. With this formulation, the statistical analysis of 4D surfaces can be cast as the problem of analyzing trajectories, or 1D curves, embedded in a nonlinear Riemannian manifold. However, performing the spatiotemporal registration, and subsequently computing statistics, on such nonlinear spaces is not straightforward as they rely on complex nonlinear optimizations. Our core contribution is the mapping of the surfaces to the space of Square-Root Normal Fields (SRNF) where the L2 metric is equivalent to the partial elastic metric in the space of surfaces. Thus, by solving the spatial registration in the SRNF space, the problem of analyzing 4D surfaces becomes the problem of analyzing trajectories embedded in the SRNF space, which has a Euclidean structure. In this paper, we develop the building blocks that enable such analysis. These include: (1) the spatiotemporal registration of arbitrarily parameterized 4D surfaces even in the presence of large elastic deformations and large variations in their execution rates, (2) the computation of geodesics between 4D surfaces, (3) the computation of statistical summaries, such as means and modes of variation, of collections of 4D surfaces, and (4) the synthesis of random 4D surfaces. We demonstrate the utility and performance of the proposed framework using 4D facial surfaces and 4D human body shapes.

scholarly journals 3D SHAPE ACQUISITION USING AN FTP-BASED METHOD IN PRODUCT MODELING

2011 ◽  
Author(s):  
Chunsheng Yu ◽  
Lushen Wu ◽  
Qingjin Peng
Keyword(s):  
Reverse Engineering ◽  
Three Dimensional ◽  
Product Modeling ◽  
Simple Method ◽  
3D Object ◽  
3D Shape ◽  
Two Dimensional Image ◽  
Shape Acquisition ◽  
3D Data ◽  
Cad Cam

Three-dimensional (3D) shape modeling is one of the most fundamental processes in CAD/CAM systems. There is a variety of methods to build 3D shapes for product design and manufacturing. The methods include defining a 3D object using solid or feature modeling methods, or building a 3D shape of the existing object using reverse engineering techniques. Image-based shape recovery techniques provide effective tools in reverse engineering to acquire 3D data of objects. This paper reports a simple method to reconstruct a 3D object from its 2D (two-dimensional) image for the product modeling. A method based on FTP (Fourier Transform Profilometry) phase analysis is proposed to measure the 3D surface of an object. The comparison of the FTP method with other methods is discussed and the process of the FTP method is provided. The experiment shows the accuracy and speed of the method.

Experimental Investigation of Behavior of Impinging Liquid Jet in a Shallow Pool by 3D-LIF

2020 ◽  
Author(s):  
Sota Yamamura ◽  
Fumihito Kimura ◽  
Hiroyuki Yoshida ◽  
Akiko Kaneko ◽  
Yutaka Abe
Keyword(s):  
Liquid Film ◽  
3D Visualization ◽  
Liquid Jet ◽  
Water Pool ◽  
Molten Material ◽  
3D Shape ◽  
The Core ◽  
Severe Accidents ◽  
Jet Velocity ◽  
Shape Data

Abstract In some scenarios of severe accidents, the core materials melt and fall into a water pool in the lower plenum as a jet. The molten material jet is broken up, and heat transfer between molten material and coolant occurs. The aim of this study is to clarify the behavior of liquid jet falling into a shallow pool. In a previous study, it is clarified that, in a shallow pool, the jet spread radially after bottoming, and the atomization occurs with high flow velocity in a shallow pool. the detail of atomization and the spreading of the jet cannot be measured by the limitation of a 2D visualization method. In this study, a 3D-LIF method is used to obtain the detail 3D shape data of the jet. The 3D visualization of the jet is conducted. Using 3D shape data, the liquid film and the atomized droplet are measured. The initial jet velocity is selected as a parameter. As a result, following knowledge is obtained. The thickness of liquid film increases suddenly, and the radius of thin liquid flow increases with the increase of the initial jet velocity. The number of atomized droplets increases with the increase of the initial jet velocity. However, the size of the droplets are not influenced by the initial jet velocity.

Sign in / Sign up

Export Citation Format

Share Document