Cight is an implementation of a visual algorithm for teach-replay navigation of mobile robots. See this article and the Dejavu demo application for reference.
Cight is developed in C++ 03 on top of boost and OpenCV. A set of cmake scripts is provided for building.
Open a command prompt, move into either folder build/release/ or build/debug/ (depending on which build mode you want to use) and enter the following commands:
$ cmake -DCMAKE_BUILD_TYPE=$MODE ../..
$ make -j2
Where $MODE should be either Release or Debug.
Alternatively, under Bash-compatible environments you can just call the build.sh script:
$ ./build.sh
The build scripts expect to find the include files for Clarus on a folder at the same level as the project's base folder, for example:
.
|
+--+ Cight
| |
| +-- build
| +-- include
| +-- src
|
+-- Clarus
|
+-- build
+-- include
+-- src
2015-03-16
A somewhat incremental update, pushed before I start porting my code base to ROS. An alternative processing pipeline more fundamentally based on difference images, which I expect will be easier to performance-optimize later on, was implemented. The previous pipeline remains available, but I plan to retire it on a future version. Also I have decided to leave compile-time diagnostics on by default.
2015-01-26
A hopelessly belated update, but better late than never. This version includes the new image pairing method described in this report. In fact the actual implementation has progressed a little beyond what is reported here, and is currently in a state of flux, but it had been too long since I uploaded anything and I was growing weary of having no backup of this work. So there.
- Increment the cmake scripts with better dependency detection;
- Document the API and source files;
- Move the code base to C++ 11, when support is widespread enough.