Ethanfel/8-cut
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: calibrate classifier probabilities with isotonic regression

Browse Source 
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
Ethanfel
2026-04-19 14:00:38 +02:00
parent 8fb8581816
commit f5361a963e
1 changed files with 11 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
core/audio_scan.py
+11
View File
@@ -560,6 +560,17 @@ def train_classifier(video_infos: list[tuple[str, list[float], list[float]]],
clf.fit(X[train_idx], y_arr[train_idx]) clf.fit(X[train_idx], y_arr[train_idx])
_log("audio_scan: classifier trained") _log("audio_scan: classifier trained")
# Calibrate probabilities for better threshold behavior
from sklearn.calibration import CalibratedClassifierCV
min_class = min(int(n_pos), int(n_neg_sample))
if min_class >= 6:
cal_clf = CalibratedClassifierCV(clf, cv=3, method='isotonic')
cal_clf.fit(X[train_idx], y_arr[train_idx])
clf = cal_clf
_log("audio_scan: classifier calibrated (isotonic, 3-fold)")
else:
_log(f"audio_scan: skipping calibration (min class size {min_class} < 6)")
model = {"classifier": clf, "n_features": X.shape[1], model = {"classifier": clf, "n_features": X.shape[1],
"embed_model": embed_model or _DEFAULT_EMBED_MODEL} "embed_model": embed_model or _DEFAULT_EMBED_MODEL}