Appendix B - Command Cheat Sheet#
Run from the project root F:\Steganography.
Purpose |
Command |
|---|---|
Install core dependencies |
pip install numpy pillow |
Install everything (ML/plot) |
pip install -e . |
Core algorithm self-test |
python src\test_core.py |
Steganalysis self-test |
python src\test_steg.py |
False-positive regression |
python src\test_false_positive.py |
GUI smoke test |
python src\test_gui.py |
End-to-end verification |
python src\run_e2e.py |
Launch the GUI |
python src\gui.py |
Code-family/efficiency plot |
python -c “import sys; sys.path.insert(0,’src’); from efficiency import plot_code_family_and_efficiency; print(plot_code_family_and_efficiency())” |
Generate v1 dataset |
python src\make_dataset.py data\campus_jpg –out campus |
Generate v2 dataset (12 variants / 143-D) |
python src\make_dataset.py –out campus_v2 –feature-set v2 –variants all |
Generate SRM-enhanced dataset (same source) |
python src\make_dataset.py data\campus_jpg –out campus_srm –preprocess srm -j 16 |
Train/evaluate a dataset |
$env:DS_FILES = “dataset.csv”; python src\train_model.py |
Train on v2 + JPEG clean |
$env:DS_FILES = “dataset_campus_v2_jpeg.csv”; python src\train_model.py |
Single-image ML prediction |
python -c “import sys; sys.path.insert(0,’src’); from ml_predict import get_predictor; import numpy as np; from PIL import Image; print(get_predictor().predict(np.asarray(Image.open(‘img/cover.png’).convert(‘L’))))” |
Switch to 53d interpretable model |
python -c “import sys; sys.path.insert(0,’src’); from ml_predict import MLPredictor; import numpy as np; from PIL import Image; print(MLPredictor(model_path=’models/stego_classifier_v2_jpeg_lgb_51d.joblib’, clip_outliers=False).predict(np.asarray(Image.open(‘img/cover.png’).convert(‘L’))))” |
C++/Python consistency check |
python -c “import sys; sys.path.insert(0,’src’); import cppembed; cppembed.selfcheck()” |
GPU dataset generation |
python gpu\make_imageset.py |
GPU feature self-check |
python gpu\featurize_gpu.py |
GPU v2 feature self-check |
python gpu\featurize_v2_gpu.py |
GPU training |
python gpu\train_ml_gpu.py |
GPU single-image detection |
python gpu\predict_gpu.py img\cover.png |
BOSSbase multi-source (GPU) |
python gpu\make_imageset.py data\BOSSbase_1.01 –out bossbase, then python gpu\train_ml_gpu.py |
Build packages |
python -m build |
Install |
pip install nsf5stego (1.9.0 on PyPI) or grab the Windows installer/portable zip |