CXR SVM demo artifact (S6-02)
Frozen inference artifact for an academic chest X-ray classifier.
- Features: 152 handcrafted features (Histogram 64 + GLCM 60 + LBP 28) from 299x299 grayscale images.
- Model:
StandardScaler+SVC(kernel="rbf", C=10.0, gamma="scale", class_weight="balanced", probability=True, random_state=42). - Classes: COVID, Lung_Opacity, Normal, Viral Pneumonia.
- Fit only on the
trainsplit of the project manifest (14,815 images). - Artifact:
cxr_inference_model_cache.joblib(7,842,399 bytes). - SHA-256:
3e14bd570c734d9d8f39a2d340b176a40a41c9559cc2da0ef67bbd8e15735058. - scikit-learn: 1.6.1.
Intended use
Academic demonstration of a classical machine-learning pipeline. Not a medical device. Do not use for diagnosis, triage, treatment, or any clinical decision. Input images must be de-identified. The model was validated only on 299x299 images from the COVID-19 Radiography Database; other sources or preprocessing are out of distribution.
Data
Derived from the public COVID-19 Radiography Database (Kaggle: tawsifurrahman/covid19-radiography-database),
whose license is reported as copyright-authors. No images are included in this repository.
Reproducibility
See the source project for the split manifest, feature code, published metrics, and the equivalence gate (200 validation images, 200/200 label match).