Ensemble model

pyproject.toml

@@ -20,6 +20,8 @@ dependencies = [
     "seaborn (>=0.13.2,<0.14.0)",
     "onnxruntime (>=1.21.0,<2.0.0)",
     "flask-ml (>=0.2.5,<0.3.0)",
+    "joblib (>=1.5.0,<2.0.0)",
+    "xgboost (>=3.0.0,<4.0.0)",
 ]
 
 [tool.poetry]

src/age_gender_classifier_ensemble_test/format_data.py

@@ -0,0 +1,26 @@
+import pandas as pd
+
+# Load the CSV
+df = pd.read_csv('temp_output_labeled.csv')
+
+# Extract true_label (only need one per imageId, they're identical across the three models)
+true_labels = df[['imageId', 'true_label']].drop_duplicates()
+
+# Pivot the data: get label and confidence per model
+pivot_df = df.pivot_table(
+    index='imageId',
+    columns='model_name',
+    values=['label', 'confidence'],
+    aggfunc='first'
+)
+
+# Flatten the multi-level column names
+pivot_df.columns = [f'{model}_{field}' for field, model in pivot_df.columns]
+
+# Merge the true_label back in
+final_df = pivot_df.reset_index().merge(true_labels, on='imageId')
+
+# Save to CSV
+final_df.to_csv('pivoted_output.csv', index=False)
+
+print("Done! Output saved to 'pivoted_output_with_true_label.csv'")

src/age_gender_classifier_ensemble_test/model.py

@@ -0,0 +1,112 @@
+import pandas as pd
+import matplotlib.pyplot as plt
+import seaborn as sns
+from sklearn.model_selection import train_test_split
+from sklearn.metrics import classification_report, confusion_matrix
+from xgboost import XGBClassifier
+
+# Load your CSV
+df = pd.read_csv('pivoted_output.csv')  
+
+# Convert true_label into an age range
+def age_to_range(age):
+    if age <= 2:
+        return "0-2"
+    elif age <= 9:
+        return "3-9"
+    elif age <= 19:
+        return "10-19"
+    elif age <= 29:
+        return "20-29"
+    elif age <= 39:
+        return "30-39"
+    elif age <= 49:
+        return "40-49"
+    elif age <= 59:
+        return "50-59"
+    elif age <= 69:
+        return "60-69"
+    else:
+        return "more than 70"
+
+df['true_range'] = df['true_label'].apply(age_to_range)
+
+# Fix inconsistent model output labels
+def fix_label(label):
+    if label == "70-79":
+        return "more than 70"
+    return label
+
+df['age_classify_v001_label'] = df['age_classify_v001_label'].apply(fix_label)
+df['vit_age_classifier_label'] = df['vit_age_classifier_label'].apply(fix_label)
+df['fairface_classifier_label'] = df['fairface_classifier_label'].apply(fix_label)
+
+# Map all labels to integers
+label_mapping = {
+    "0-2": 0,
+    "3-9": 1,
+    "10-19": 2,
+    "20-29": 3,
+    "30-39": 4,
+    "40-49": 5,
+    "50-59": 6,
+    "60-69": 7,
+    "more than 70": 8
+}
+df['age_classify_v001_label'] = df['age_classify_v001_label'].map(label_mapping)
+df['vit_age_classifier_label'] = df['vit_age_classifier_label'].map(label_mapping)
+df['fairface_classifier_label'] = df['fairface_classifier_label'].map(label_mapping)
+
+# Feature and target setup
+X = df[
+    [
+        'age_classify_v001_label', 'age_classify_v001_confidence',
+        'vit_age_classifier_label', 'vit_age_classifier_confidence',
+        'fairface_classifier_label', 'fairface_classifier_confidence'
+    ]
+]
+y = df['true_range'].map(label_mapping)
+
+# Train/test split
+X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
+
+# Final XGBoost model using the best hyperparameters
+clf = XGBClassifier(
+    random_state=42,
+    use_label_encoder=False,
+    eval_metric='mlogloss',
+    colsample_bytree=0.6,
+    gamma=0,
+    learning_rate=0.01,
+    max_depth=3,
+    n_estimators=500,
+    subsample=0.6
+)
+
+# Train
+clf.fit(X_train, y_train)
+
+# Predict and evaluate
+y_pred = clf.predict(X_test)
+print(classification_report(y_test, y_pred))
+
+# ---- Plot Confusion Matrix ----
+
+# Compute confusion matrix
+cm = confusion_matrix(y_test, y_pred)
+
+# Define readable class labels
+class_labels = [
+    "0-2", "3-9", "10-19", "20-29", "30-39", 
+    "40-49", "50-59", "60-69", "70+"
+]
+
+# Plot
+plt.figure(figsize=(10, 8))
+sns.heatmap(cm, annot=True, fmt='d', cmap='Blues',
+            xticklabels=class_labels, yticklabels=class_labels)
+
+plt.xlabel('Predicted')
+plt.ylabel('True')
+plt.title('Confusion Matrix')
+plt.show()