BlacknDecker/ImageFailureInjection

Image Failure Injection

This code collects some tools to inject image failures and tries to simplify the usage for dataset injection.

The work is based on:

• https://github.com/AndreaCeccarelli/Python_Image_Failures
• https://github.com/UjjwalSaxena/Automold--Road-Augmentation-Library

The InjectionManager takes as an input a folder with a sequence of frames (png images) and injects the selected failures. The output is a folder (named as the initial sequence) with a subfolder for every injection. Every failure could have multiple variants, therefore inside the failure output folder a new subfolder is created for every variant. Every failure variant folder contains the initial sequence with the injected failure.

Example:

• input
  • sequence_01
    • frame_01.png
    • frame_02.png
    • ...

If we inject "banding" and "condensation" failures the output would be:

• output
  • sequence_01
    • banding
      • variant_0
        • frame_01.png
        • frame_02.png
        • ...
      • variant_1
        • frame_01.png
        • frame_02.png
        • ...
    • condensation
      • variant_0
        • ...