SimCATS-Datasets v2.6.0

SimCATS-Datasets is a Python package that simplifies the creation and loading of SimCATS datasets. Please have a look at this repository regarding SimCATS itself.

Please have a look at the documentation on ReadTheDocs

Changes in this version:

• Full support for sensor scan datasets introduced in SimCATS 2.0
• Dataset creation:
  • create_dataset now supports the creation of sensor scan datasets.
  • create_simulated_dataset now supports the creation of sensor scan datasets. This also includes new functionality that automatically computes additional label masks from the metadata (pinch-off label masks), which have been implemented by Karin Havemann during the course of her master thesis.
  • create_simulated_dataset now offers the option to reset the sensor potential offset if creating a CSD dataset. This effectively enables simulation of retuning the sensor to the selected working point before each CSD measurement.
  • create_simulated_dataset now supports the creation of datasets consisting of one-dimensional measurements.
• Data preprocessing:
  • quantize_to_int_bit_depth allows quantization of the data to different integer precisions. This effectively enables testing the effect of different data precisions for efficiency analysis.
  • resample_image_to_32x32 allows to use the resampling to 32x32 preprocessor in the Pytorch Dataset class by specifying it via a string.
  • min_max_0_1_given_limits allows scaling data to the range 0 to 1 while taking into account given minimum and maximum values. This ensures consistent scaling across multiple individual datasets.
  • min_max_0_1_global scales a full dataset to the range 0 to 1. The legacy function min_max_0_1 only scales each individual data sample to 0 to 1, ignoring the global range.
• Added several helper functions and new ground truth types for sensor scan datasets. Please refer to the documentation for a description of the available ground truth types.
• Bug fixes:
  • clip_point_line_to_rectangle returned incorrect coordinates when the start or end point were exactly on the border of the rectangle.