CaveGen uses a pseudo-randomized, Perlin noise generated, monochrome noisemaps and a box-blur convolution kernel, which are being pulled through an OpenCV binary threshold algorithm. For examples, see below. The carving is implemented through matrix' depth-first search, always starting from the cave's center coordinates.
CaveGen can be built from source using either g++ or cl. As of pre-alpha, CaveGen is platform-independent, meaning it can be built both on Windows and other UNIX-like systems. Future plans include an implementation of Windows' API. Minimum required C++ standard is C++20.
Last revision: 0.1a
In order to compile without linking, run: cl /EHsc /c /std:c++20 *.cpp . After the compilation process finished successfully, run: cl *.obj /out:{filename} to link compiled object files, where filename is the linked output (executable) file.
Note: If using Visual Studio Code as your preferred IDE, remember that it needs to be launched using
Windows' Developer Command Prompt for VS, in order to acknowledge the included Windows API libraries, and for the cl run command to be recognized.
Compiling via the GNU's Conpiler Collection is fairly straightforward - simply run g++ *.cpp -o {filename}, and you should be presented with a working, aptly-named executable file.
Since CaveGen is presented via a CLI narrative, most of your processing will be done through the command line. There are primitively-supported program parameters:
width: integer, width-dimension in pixels of the resulting cave imageheight: integer, height-dimension in pixels of the resulting cave imagecarving-threshold: float, minimum value of the (x, y) coordinates on the noisemap for the carving algorithm
The program can be invoked with the following syntax:
cavegen.exe [width] [height] [carving-threshold]
Both width and height are interchangeable, meaning you can generate a square matrix providing only one of the width or height arguments: cavegen [length] [carving-threshold], but the carving threshold can not be left out.
As a result of your pseudo-random luck, you will be presented with original_cave.bmp bitmap, located within the ./examples/ folder. If you are not satisfied with the result, and depending whether the original output was box-blurred (or not), you can run the thresholding.py program (Usage: see below), to generate a more natural, 2D cave system.
To use the Python thresholding module, make sure you have installed the prerequisites, listed within the requirements.txt file, using the pip install -r requirements.txt command. After the installation has been successful, you can run the Python file using the following syntax:
python thresholding.py --filename [string] --lower [int, 0-255] --upper [int, lower-255]
which will produce two resulting images in the examples folder:
adaptive_cave.bmp- result of the adaptive thresholding (from OpenCV's library)binary_cave.bmp- result of the barebones binary thresholding (also from OpenCV's library)
Note: Future plans include implementation of advanced command-line arguments' processing via Boost C++ Libraries (boost/program_options), so as of 0.1a (pre-alpha), only the primitive support is available.
Important note: Please respect the implemented randomness of the modified Perlin's algorithm, so if you get a totally black cave - the center coordinates are equal to zero, please test your luck and try again, or play with the provided criteria :)
| Attributes | Preview |
|---|---|
| Unmodified cave, criteria=0.36, blur=1.0: |  |
| Binary thresholded, lower=85, upper=255: |  |
| Adaptive thresholded, lower=85, upper=255, 11:2, mean |  |
| Attributes | Preview |
|---|---|
| Unmodified cave, criteria=0.37, blur=1.0: |  |
| Binary thresholded, lower=85, upper=255: |  |
| Adaptive thresholded, lower=85, upper=255, 11:2, mean |  |
| Attributes | Preview |
|---|---|
| Unmodified cave, criteria=0.3667, blur=1.0: |  |
| Binary thresholded, lower=20, upper=255: |  |
| Adaptive thresholded, lower=20, upper=255, 11:2, mean |  |
I have to give credit where credit is due:
- Bitmap library from kbuffardi, made my life writing/reading bitmaps way easier.
- Ken Perlin's noise algorithm from this wiki.