Every point matters!
The main goal of this project is to design and implement methods for 3D Scene Reconstruction, automatic generation of simple semantic 3D data description obtained from stereo reconstruction and the Hierarchical 3D Stitching of surface patches. The outcome of the implementation will be a working prototype for 3D data reconstruction.
As an input, we are using freely available point cloud scans.We are focusing on the PCD, LAS as well as XYZ CSV point cloud data formats. Namely, we are using datasets created by Autonomous Systems Lab of Institute of Robotics and Intelligent Systems at ETH Zurich.
As a collaboration platform, we are using the Microsoft's Team Foundation Server. It has built-in functionality for agile project management, source code management, task management and many others. Its native integration with Visual Studio made us to decide for choosing this management tool.
Output of our effort will be a tool for point cloud data reconstruction into abstract 3D geometric shapes model. It will also take care of hierarchical registration and surface stitching for multiple view images. Compared to raw point cloud data, such a parametric model will be far more convenient to further editing or physical reproduction, e.g. on a 3D printer. This tool will also provide basic visualization for viewing both input and reconstructed output.
As a collaboration platform, we are using the Microsoft's Team Foundation Server. It has built-in functionality for agile project management, source code management, task management and many others. Its native integration with Visual Studio made us to decide for choosing this management tool.
Using the hottest cutting edge technologies, we are going to bring you the most state-of-the-art reverse engineering 3D reconstruction tool ever made. We are trying our best to pick the most feasible means from various image and 3D data processing libraries, to develop a tool for point cloud data reconstruction into abstract 3D geometric shapes model.
Our wast experience in computer vision combined with high motivation gives us a good predisposition for a successful project. We promise to do our best pushing each other forward to our goal.
Architecture manager (Team leader)
He obtained Bachelors degree by completing a field of study at the FIIT STU. With the final work he got familiar with computer vision and was so enthused that he dared to continue in this field in Team project...
Until now he had experience especially with the Java programming language, C and thanks to his final thesis, he moved to C++. In addition, however, he is experienced in database technology, CPU paralell programming and basics in Computer Graphics.
Development manager (Deputy Leader)
He is a graduate of bachelor studies at FIIT STU in department of Informatics. During his studies he was dealing with Image segmentation using graphics processing unit, particulary with superpixels segmentation and their sequential merging...
He published his work on conferences IIT.SRC 2015 and CESCG 2015. He has experience with several programming languages and technologies (C/C++, Java, CUDA, OpenCV, OpenGL, MPI, SQL, MySQL, ABAP). Currently he is interested in computer vision and computer graphics.
Test manager
Young brave programmer interested in node.js and Lua. Now working so hard in C++ and wants to help the team as much as possible. His main work is 3D reconstruction. When he doesn't code he probably trains or party.
Business manager
Graduate of the Informatics study program at FIIT STU. In his bachelor thesis, he was utilizing the Yang modeling language for remote management of OpenVZ virtualization platform. During his studies, he got in touch with C, Java, PostgreSQL and Python...
He is passionate about digging in pretty much any system. In his diploma thesis, he is going to deal with Intrusion detection by machine learning. Likes to play table tennis in his free time.
Documentation manager
A graduate of bachelor studies at FIIT STU, study program Informatics in 2015. The main topic of her bachelor thesis was Time series analysis, specifically various methods and models for time series prediction, in the field of energetics...
During her studies she gained experience with several programming languages such as C, Java, PHP, Ruby on Rails and begins with C ++. In this time, she is a student of Master degree at FIIT (study program: Information Systems).
Programmer
He is a graduate of Bachelor studies at FIIT STU in the Department of Informatics, during the study dealt with programming support for OpenCL applications on the GPU. He has experience with programming languages C, C ++, Java, HTML...
...MySQL database system, Cisco network technology and basics of graphics. At present, he continues to expand his knowledge at FIIT STU in the study of information systems.
Propagation manager
A smily guy with good sense for order. Also graduated in Informatics from FIIT STU. His bachelor thesis entitled "Household classification and visualization based on electricity consumption" involved data analytics, statistics and graph visualization...
Therefrom, he got familiar with the R language for statistical computing. But now he is more interested in web technologies. He loves to drive and to take care of his car.
Our final software solution is very interesting and allows easy interaction and visualization for architects and interior designers, whom we simplify and speed up the work plans. Here are two main vizualization types our solution offers:
Four different views of data are supported - original data cloud, recognized border points, reconstructed shapes, remaining objects (in development). Border points of each segment have their own color and 10px size to distinguish from original data and displayed concave hulls. Reconstructed shapes are distinguished with different colors - Color Generator and pseudo random selector.
Our tool allows automatic blueprint creation of desired scene, so architects can be instantly provided with its simplified 3D CAD model. As a output format the widely supported DXF format is used. When such a model is opened in a CAD tool, i.e. AutoCAD, it can be further edited and exact dimensions of various objects can be measured. The export process is done using sgCore solid geometry library. Currently, export of planar objects and polygon mesh is supported. In future, adding objects created by boolean operations (i.e. intersection of a cylinder and a sphere) can be implemented easily, when needed.
Every thursday at 10:00 AM, we are putting our heads together in the room 1.27 (Laboratory of Vision and Graphics Group), to evaluate our progress and to discuss all the newly arising problems. Here you can find weekly reports from these meetings.
