Documentation

• Feature collection
• Parameter settings
• Image segmentation
• Troubleshooting

Feature collection

• Configurations:
  • Output prefix: prefix for output files; default: feature-collection-.
  • Image filename pattern: image filename pattern matching rules, split by space; default: *.tiff *.png. See gnu.org.
  • Image path: path to the image files
  • Output path: path for output files
• Features: select the features to be collected
  • Area features
  • Spatial features
  • Haralick features
• Commands:
  • Start: start feature collection
  • Stop: stop feature collection; collected features would not be saved
  • Parameters: edit parameter settings
  • Documentation: open documentation page in default browser
  • Close: close the window

Parameter settings

• Parameters
  • distance (int, default=1): the distance parameter from mahotas.features.haralick(); see link
  • P (int, default=10): the P parameter from skimage.feature.local_binary_pattern(); see link
  • R (float, default=5): the R parameter from skimage.feature.local_binary_pattern(); see link
  • d (int, default=15): the d parameter from cv2.bilateralFilter(); see link
  • sigma_color (double, default=15): the sigma_color parameter from cv2.bilateralFilter(); see link
  • sigma_space (double, default=15): the sigma_space parameter from cv2.bilateralFilter(); see link
• Closing then opening kernel size (k x k)
  • Closing k (int, default=9)
  • Opening k (int, default=9)
• Opening then closing kernel size (k x k)
  • Opening k (int, default=9)
  • Closing k (int, default=3)
• Commands:
  • Submit: save the parameters
  • Documentation: open documentation page in default browser
  • Close: close the window

Image segmentation

Configurations:

• Image filename: image filename
• Output path: path for output files

Troubleshooting

1. The program crashed? Unfortunately, this software is not exhaustively tested for boundary cases. Unexpected inputs may crash the program. If you think this is a bug, feel free to send me a bug report!

2. My image is cropped? Following the design of our dataset, most images come with a margin describing the camera and sample parameters. Input images of the following sizes will be cropped automatically: 2048x2560, 1428x2048, 1024x1280, 1448x2048.

Last update: Oct. 05, 2020