feat: enhance visualization pipeline with additional figures and manifest support

app/training/visualize_results.py

@@ -945,6 +945,7 @@ def generate_all_figures(
     results_path: str | Path,
     features_csv: str | Path,
     output_dir: str | Path = "figures",
+    manifest_csv: str | Path | None = None,
 ) -> None:
     """Generate all publication-quality figures."""
     results_path = Path(results_path)
@@ -952,47 +953,70 @@ def generate_all_figures(
     output_dir = Path(output_dir)
     output_dir.mkdir(parents=True, exist_ok=True)
 
-    print(f"\nLoading results from {results_path}...")
+    if manifest_csv is None:
+        manifest_csv = features_csv.parent / "manifest.csv"
+    manifest_csv = Path(manifest_csv)
+
+    print(f"\nLoading results from {results_path}...", flush=True)
     with open(results_path, "r") as f:
         saved_results = json.load(f)
 
-    # We need y_true/y_pred/y_prob arrays — check if they're stored
-    # If not in the saved JSON, we can't plot ROC/PR curves
     has_arrays = any(
         "y_true" in v for k, v in saved_results.items()
         if isinstance(v, dict) and not k.startswith("_")
     )
 
-    print(f"\nGenerating figures in {output_dir}/...\n")
+    print(f"\nGenerating figures in {output_dir}/...\n", flush=True)
 
     if has_arrays:
-        print("[1/8] ROC Curves")
+        print("[1/13] ROC Curves", flush=True)
         plot_roc_curves(saved_results, output_dir)
 
-        print("[2/8] Precision-Recall Curves")
+        print("[2/13] Precision-Recall Curves", flush=True)
         plot_pr_curves(saved_results, output_dir)
 
-        print("[3/8] Confusion Matrices")
+        print("[3/13] Confusion Matrices", flush=True)
         plot_confusion_matrices(saved_results, output_dir)
     else:
-        print("[1-3/8] Skipping ROC/PR/CM — no per-sample predictions stored")
+        print("[1-3/13] Skipping ROC/PR/CM — no per-sample predictions stored", flush=True)
 
-    print("[4/8] Model Comparison Bar Chart")
+    print("[4/13] Model Comparison Bar Chart", flush=True)
     plot_model_comparison(saved_results, output_dir)
 
-    print("[5/8] Feature Importance")
+    print("[5/13] Feature Importance", flush=True)
     plot_feature_importance(saved_results, output_dir)
 
-    print("[6/8] Feature Correlation Heatmap")
+    print("[6/13] Feature Correlation Heatmap", flush=True)
     plot_correlation_heatmap(features_csv, output_dir)
 
-    print("[7/8] Feature Distributions (AI vs Human)")
+    print("[7/13] Feature Distributions (AI vs Human)", flush=True)
     plot_feature_distributions(features_csv, output_dir, saved_results)
 
-    print("[8/8] LaTeX / Markdown Tables")
+    print("[8/13] LaTeX / Markdown Tables", flush=True)
     generate_latex_table(saved_results, output_dir)
 
-    print(f"\nAll figures saved to {output_dir}/")
+    print("[9/13] Dataset Statistics", flush=True)
+    plot_dataset_statistics(manifest_csv, output_dir)
+
+    print("[10/13] Per-Source Class Balance", flush=True)
+    plot_per_source_balance(manifest_csv, output_dir)
+
+    print("[11/13] Feature Space Embedding (PCA + t-SNE)", flush=True)
+    plot_feature_embedding(features_csv, output_dir)
+
+    if has_arrays:
+        print("[12/13] Calibration Curves", flush=True)
+        plot_calibration_curves(saved_results, output_dir)
+
+        print("[13/13] Prediction Score Distribution", flush=True)
+        plot_score_distribution(saved_results, output_dir)
+    else:
+        print("[12-13/13] Skipping calibration/score dist — no probabilities stored", flush=True)
+
+    print("\n[EXTRA] Report Summary (Markdown)", flush=True)
+    generate_extended_summary(saved_results, manifest_csv, features_csv, output_dir)
+
+    print(f"\nAll figures saved to {output_dir}/", flush=True)
 
 
 if __name__ == "__main__":