How gbrlfaria/rune-breaker works
Project Type & Purpose
• Its purpose is to test and demonstrate how modern ML techniques can bypass visual obfuscation in anti-botting mechanisms.
• The pipeline covers multiple phases—from data collection & labeling to preprocessing, dataset generation, CNN model training, and evaluation.
Breaking Down the File Structure into Main Components
• Core Files and Utilities:
– “common.py” contains general-purpose functions possibly shared across modules.
– Standard configuration files (e.g. .gitignore, LICENSE, requirements.txt) support development.
• Preprocessing Tools:
– In the “preprocessing” folder, “label.py” is the interactive labeling tool.
• Operations for Dataset Generation:
– “operations/make_dataset.py” creates balanced, augmented training, validation, and testing splits.
– “operations/revert_dataset.py” (likely to reverse/adjust dataset processing if needed).
• Model Implementation:
– “model/train.py” sets up and trains the CNN architecture using Keras.
– “model/classify.py” runs evaluation/classification on the testing set.
• Results:
– “results” folder holds output logs and experiment summaries (e.g. denoising.txt, grayscale.txt, pipeline.txt, thresholding.txt).