🔍 Header/Footer Detection Explorer

Loading...
⚙️ Confidence Filters (click to expand)

Set minimum confidence thresholds to hide low-confidence results. HYBRID method is never filtered.

0.50
0.40
0.50
0.55
0.60
0.45
0.55

📂 Processed Files (0)

No processed files yet. Select a PDF and click Analyze.

Method Comparison

Confidence Distribution

Position Analysis

⚙️ Detection Configuration

Adjust detection parameters to fine-tune header/footer identification. Changes will be applied when you analyze a document.

🌍 Global Settings

Global maximum Y-position for headers (default for all groups)
Global minimum Y-position for footers (default for all groups)

🎯 Confidence Groups (Multi-Strategy Detection)

📊 Group 1: DEFAULT_GROUP (Weighted Voting)
Methods: All 7 methods (Z-score, IQR, Pattern, Layout, Font, NER, PageConsistency)
Logic: Sum weights of agreeing methods, accept if total ≥ threshold
Minimum weighted score (max 21.0 with current weights)
Minimum weighted average confidence
🔒 Group 2: STRICT_GROUP (AND Gate)
Methods: Z-score, IQR, Pattern (3 core methods)
Logic: Accept ONLY if ALL 3 methods agree
Minimum average confidence (all 3 methods must be confident)

⚖️ Method Weights (Default total: 8.00 - from 0.00 to 3.00)

▶ Method-Specific Parameters

💡 Tip: Higher weights give more importance to specific methods. Increase min_weight for stricter detection, decrease for more permissive results.

✓ Manual Annotations

Headers

Footers

📊 Feedback Summary

Method Accuracy

Metrics Scope:

🤖 Machine Learning Metrics

Click "Calculate ML Metrics" button above to generate comprehensive evaluation metrics.

Loading available files...