TeleSculptor is a cross-platform desktop application for photogrammetry. It was designed with a focus on aerial video, such as video collected from UAVs, and handles geospatial coordinates and can make use of metadata, if available, from GPS and IMU sensors. However, the software can also work with non-geospatial data and with collections of images instead of metadata. TeleSculptor uses Structure-from-Motion techniques to estimate camera parameters as well as a sparse set of 3D landmarks. It uses Multiview Stereo techniques to estimate dense depth maps on key frame and then fuses those depth maps into a consistent surface mesh which can be colored from the source imagery.
TeleSculptor can be installed from precompiled binaries for Linux, MacOS, and Windows included at the bottom of the latest release page by following the instructions in the Installation section. Instructions on how to use the TeleSculptor GUI can be found in the User Guide. A computer with at least 16GB of RAM is recommended for processing most datasets.
More advanced users who wish to build the project from source should proceed to the Building TeleSculptor section.
TeleSculptor provides a graphical user interface with Qt, 3D visualization with VTK, and photogrammetry algorithms with KWIVER. This project was previously called MAP-Tk (Motion-imagery Aerial Photogrammetry Toolkit). The MAP-Tk name is still scattered throughout the source code. MAP-Tk started as an open source C++ collection of libraries and tools for making measurements from aerial video. The TeleSculptor application was added to the project later. The original software framework and algorithm were then refactored into KWIVER and then expanded to address broader computer vision problems. While KWIVER is now a more broad set of tools, TeleSculptor remains an application focused on photogrammetry.
The advantage of the KWIVER software architecture (previously MAP-Tk) is that it is highly modular and provides an algorithm abstraction layer that allows seamless interchange and run-time selection of algorithms from various other open source projects like OpenCV, VXL, Ceres Solver, and PROJ4. The core KWIVER library (vital) and tools are light-weight with minimal dependencies (C++ standard library, and Eigen). TeleSculptor is written to depend only on the KWIVER "vital" library. Additional capabilities are provided by KWIVER arrows (plugin modules) that use third party libraries to implement various abstract algorithm interfaces defined in the KWIVER vital library. This means that new plugins can be dropped into TeleSculptor to enable alternative or new functionality by adjusting some settings in a configuration file. While TeleSculptor provides a default workflow that works out of the box, it is not just an end user tool. It is designed to be highly configurable to support research into to algorithms and new problem domains.
The screenshots below show TeleSculptor running on example videos from the VIRAT Video Dataset, CLIF 2007 Dataset, and other public data sets. More information about this example data can be found in the examples directory.
While the initial software implementation relies on batch post-processing of aerial video, our intent is to move to an online video stream processing framework and optimize the algorithm to run in real-time.
If you have downloaded an installer from the latest release you can simply install TeleSculptor according to the instructions for your operating system described below. If you are building TeleSculptor from source you should proceed to Building TeleSculptor to create the installer before completing the installation.
Windows: run the installer executable (exe) and follow the prompts in the installer dialog. Administrative permission is required.
Mac: open the disk image (dmg), accept the license terms, then drag the TeleSculptor application into the Applications folder.
Linux: open a bash/cmd shell and run the self extracting installer script (sh). You can view additional installation options using ./TeleSculptor-<version>-Linux-x86_64.sh --help
The remainder of this document is aimed at developers who wish to build the project from source. For end users looking for instruction on running the GUI application please read the User Guide.
TeleSculptor requires C++11 compliant compiler (e.g. GCC 4.8.1, Clang 3.3, Visual Studio 2015). TeleSculptor uses CMake (www.cmake.org) for easy cross-platform compilation. The minimum required version of CMake is 3.9.5, but newer versions are recommended.
The build is directed by CMake to ensure it can be built on various platforms. The code is built by a CMake 'superbuild', meaning as part of the build, CMake will download and build any dependent libraries needed by TeleSculptor. The build is also out of source, meaning the code base is to be separate from the build files. This means you will need two folders, one for the source code and one for the build files. Here is the quickest way to build via a cmd/bash shell.
Before building on Linux systems you must install the following packages:
On Linux, to optionally build with Python and to build the user documentation, you will also need to install the following:
Set up the folder structure and obtain the source files. This can be done with git or by downloading the files and extracting them. Then setup the folder(s) to build the binary files.
Generate the makefile/msvc solution to perform the superbuild using cmake. A description of the configuration options can be found in CMake Options.
Build the installer target/project
You have now built a TeleSculptor installer similar to what is provided in the latest release section. To install TeleSculptor on you machine, follow the instructions above in Installation.
CMAKE_BUILD_TYPE |
The compiler mode, usually Debug or Release |
TELESCULPTOR_ENABLE_CUDA |
Enable GPU acceleration with CUDA |
TELESCULPTOR_ENABLE_PYTHON |
Enable Python bindings in KWIVER |
TELESCULPTOR_ENABLE_MANUALS |
Turn on building the user documentation |
TELESCULPTOR_ENABLE_TESTING |
Build the unit tests |
TELESCULPTOR_SUPERBUILD |
Build as a superbuild (build Fletch and KWIVER) |
By default the CMAKE_BUILD_TYPE is set to Release.
Separate directories are required for Debug and Release builds, requiring CMake to be run for each.
Even if you are using a Multi-Configuration build tool (like MSVC) to build Debug you must select the Debug CMAKE_BUILD_TYPE. (On Windows in order to debug a project all dependent projects must be build with Debug information.)
For MSVC users wanting a RelWithDebInfo build we recommend you still choose Release for the superbuild. Release and RelWithDebInfo are compatible with each other, and Fletch will build its base libraries as Release. MSVC solutions will provide both Release and RelWithDebInfo configuration options. You will need to open the <build/directory>/external/kwiver-build/KWIVER.sln and build this solution with the RelWithDebInfo configuration.
The TeleSculptor GUI application is enabled by default, and all dependencies will be built by the Superbuild.
If TELESCULPTOR_ENABLE_MANUALS is enabled, and CMake finds all dependencies, then the user manuals are built as part of the normal build process under the target "manuals". The GUI manual can be viewed from inside the GUI by choosing the "TeleSculptor User Manual" action from the "Help" menu.
To build the user manual(s), you need:
- Python
version 3.4 or greater http://www.python.org/
- Docutils
version 0.11 or greater http://docutils.sourceforge.net/
(At present, only the GUI has a user manual. Other manuals may be added in the future.)
Continuous integration testing is provided by CDash. Our MAP-Tk dashboard hosts nightly build and test results across multiple platforms including Windows, Mac, and Linux.
Anyone can contribute a build to this dashboard using the dashboard script provided. Follow the instructions in the comments.
Travis CI is also used for continued integration testing. Travis CI is limited to a single platform (Ubuntu Linux), but provides automated testing of all topic branches and pull requests whenever they are created.
| Travis CI master branch: |  _ |
| Travis CI release branch: |  _ |
TeleSculptor is built on top of the KWIVER toolkit, which is in turn built on the Fletch super build system. As mentioned above, to make it easier to build TeleSculptor, a "super-build" is provided to build both KWIVER and Fletch. But, if you wish, you may point the TeleSculptor build to use your own KWIVER builds.
If you would like TeleSculptor to use a prebuilt version of KWIVER, specify the kwiver_DIR flag to CMake. The kwiver_DIR is the KWIVER build directory root, which contains the kwiver-config.cmake file.
You must ensure that the specified build of KWIVER was built with at least the following options set:
The required KWIVER flags can be found in this file : CMake/telesculptor-external-kwiver.cmake
The required Fletch flags can be found in this file : CMake/telesculptor-external-fletch.cmake
CMake |
contains CMake helper scripts |
config |
contains reusable default algorithm configuration files |
doc |
contains release notes, manuals, and other documentation |
examples |
contains pointers to example public datasets to use |
gui |
contains the visualization GUI source code and headers |
gui/icons |
contains the visualization GUI icon resources |
maptk |
contains the maptk library source and headers |
packaging |
contains support files for CPack packaging |
scripts |
contains Python helper scripts |
plugins/blender |
contains Python plug-ins for Blender |
plugins/sketchup |
contains Ruby plug-ins for SketchUp |
tests |
contains testing framework and tests for each module |
TeleSculptor is a component of Kitware's collection of open source computer vision tools and part of the KWIVER ecosystem. Please join the kwiver-users mailing list to discuss or to ask for help with using TeleSculptor. For less frequent announcements about TeleSculptor and other KWIVER components, please join the kwiver-announce mailing list.
The authors would like to thank AFRL/Sensors Directorate for their support of this work via SBIR Contract FA8650-14-C-1820. This document is approved for public release via 88ABW-2015-2555.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent