This page is a description of the tools used along with Training ARToolKit Natural Feature Tracking (NFT) to Recognize and Track an Image.
The checkResolution tool supplied with ARToolKit can help in determining the required resolution of source image data used in creating an NFT dataset.
- Mac OS X/Linux: `./checkResolution` - Windows: `checkResolution.exe
You will be prompted to enter the size of the Hiro marker. E.g. if printed at 40 mm size, enter
Moving the camera around and observing the DPI values should give you an idea of the Training ARToolKit Natural Feature Tracking (NFT) to Recognize and Track an Image required when producing the digital version of the printed material to be tracked (it is not recommended to produce imagery at a higher resolution than the printed version, which is typically 150dpi). Additionally, the output helps determine the range of resolutions required when running the genImageSet tool as the first step in training a new NFT data set.
Be sure to use a camera running at the same frame size as will be used in the online tracking process; the DPI values produced depend on the camera image size. In spite of megapixel webcams being the norm, it is actually better to use a lower resolution camera with a higher frame rate; 640×480 is perfectly adequate for most NFT tracking situations.
Below is a table of keyboard / mouse controls for using checkResolution:
|1||Decrease binarization threshold|
|2||Increase binarization threshold.|
dispFeatureSet displays trained NFT datasets by overlaying representations of the data points on the source images.
./dispFeatureSet <filename> -fset Show fset features. -fset3 Show fset3 features.
After launching dispFeatureSet, the various image resolutions will be displayed on screen with the tracking features overlaid. The features used in continuous tracking are outlined by red boxes, and the features used in identifying the pages and initializing tracking are marked by green crosses.
dispImageSet displays compressed image pyramids.
After launching dispImageSet, the various image resolutions will be displayed on screen (shrunk/zoomed as necessary to fit on screen). Press spacebar to view the images, or esc when you're done.
genTexData performs training of NFT datasets from a supplied JPEG-format source image.
./genTexData <filename> -level=n (n is an integer in range 0 (few) to 4 (many). Default 2.' -sd_thresh=<sd_thresh> -max_thresh=<max_thresh> -min_thresh=<min_thresh> -leveli=n (n is an integer in range 0 (few) to 3 (many). Default 1.' -feature_density=<feature_density> -dpi=<dpi> -max_dpi=<max_dpi> -min_dpi=<min_dpi> -background Run in background, i.e. as daemon detached from controlling terminal. (Mac OS X and Linux only.) -log=<path> -loglevel=x x is one of: DEBUG, INFO, WARN, ERROR. Default is INFO. -exitcode=<path> --help -h -? Display this help
E_NO_ERROR = 0 E_BAD_PARAMETER = 64 E_INPUT_DATA_ERROR = 65 E_USER_INPUT_CANCELLED = 66 E_BACKGROUND_OPERATION_UNSUPPORTED = 69 E_DATA_PROCESSING_ERROR = 70 E_UNABLE_TO_DETACH_FROM_CONTROLLING_TERMINAL = 71 E_GENERIC_ERROR = 255
See Training ARToolKit Natural Feature Tracking (NFT) to Recognize and Track an Image for more information on NFT datasets.
Last modified: 2016/02/15 05:41 (external edit)