"""
Analyze 'sample video.mp4' through the SimShieldAI suspicious behavior pipeline.
Prints detailed per-frame diagnostics for violence and running detection.
"""
import os
import sys
import time
import cv2
import numpy as np

sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))

from suspicious_behavior.pipeline.frame_analyzer import FrameAnalyzer
from suspicious_behavior.pipeline.video_processor import VideoProcessor


def main():
    video_path = r"C:\Users\Admin\Downloads\sample video.mp4"
    
    print("=" * 70)
    print("SimShieldAI — Suspicious Behavior Detection Analysis")
    print(f"Video: {video_path}")
    print("=" * 70)
    
    # Get video metadata
    processor = VideoProcessor(target_fps=10.0)
    metadata = processor.get_metadata(video_path)
    print(f"Resolution: {metadata['width']}x{metadata['height']}")
    print(f"Duration: {metadata['duration']:.1f}s | Frames: {metadata['frame_count']} @ {metadata['fps']:.1f} FPS")
    print(f"Processing at 10 FPS (every {int(metadata['fps']/10)} frames)")
    print("=" * 70)
    
    # Initialize analyzer
    print("\n[Init] Loading models (YOLOv8n + VideoMAE + MediaPipe)...")
    analyzer = FrameAnalyzer(camera_id="test_cam", violence_stride=4)
    
    # Reset state
    analyzer.frame_idx = 0
    analyzer.violence_buffer.clear()
    analyzer.alert_manager.reset_cooldowns()
    
    # Process video
    frames_processed = 0
    violence_scores = []
    running_detections = []
    all_alerts = []
    output_frames = []
    
    start_time = time.time()
    
    print("\n--- Per-Frame Analysis ---\n")
    
    for f_idx, frame, timestamp in processor.extract_frames_generator(video_path):
        annotated, frame_alerts, frame_meta = analyzer.analyze(frame, fps=10.0, output_base64=False)
        frames_processed += 1
        output_frames.append(annotated)
        
        # Collect metrics
        v_meta = frame_meta["violence_metrics"]
        if v_meta["confidence"] > 0:
            violence_scores.append({
                "frame": f_idx,
                "time": timestamp,
                "label": v_meta["label"],
                "conf": v_meta["confidence"]
            })
        
        for track in frame_meta["tracks"]:
            running_detections.append({
                "frame": f_idx,
                "time": timestamp,
                "track_id": track["track_id"],
                "behavior": track["behavior"],
                "conf": track["behavior_confidence"]
            })
        
        for alert in frame_alerts:
            all_alerts.append({"time": timestamp, "alert": alert})
        
        # Print frame details
        tracks_info = ", ".join([
            f"ID{t['track_id']}:{t['behavior']}({t['behavior_confidence']*100:.0f}%)" 
            for t in frame_meta["tracks"]
        ]) or "no people"
        
        violence_info = ""
        if v_meta['confidence'] > 0:
            emoji = "🔴" if v_meta['detected'] else "🟢"
            violence_info = f" | {emoji} Violence: {v_meta['label']}({v_meta['confidence']*100:.1f}%)"
        
        alert_info = ""
        if frame_meta['threats_detected']:
            alert_info = f" | ⚠️ THREATS: {frame_meta['threats_detected']}"
        
        print(f"  [{f_idx:04d}] {timestamp:5.1f}s | {frame_meta['active_tracks_count']} people | "
              f"[{tracks_info}]{violence_info}{alert_info}")
    
    elapsed = time.time() - start_time
    
    # Save annotated output
    output_path = r"C:\Users\Admin\Downloads\sample_video_annotated.mp4"
    processor.write_frames_to_video(
        output_frames, output_path, fps=10.0,
        frame_size=(metadata['width'], metadata['height'])
    )
    
    # Print summary
    print("\n" + "=" * 70)
    print("ANALYSIS SUMMARY")
    print("=" * 70)
    print(f"Frames processed: {frames_processed}")
    print(f"Processing time: {elapsed:.1f}s ({frames_processed/elapsed:.1f} FPS)")
    
    print(f"\n--- Violence Detection (VideoMAE) ---")
    if violence_scores:
        confs = [s["conf"] for s in violence_scores]
        violence_detected = [s for s in violence_scores if s["label"] != "normal"]
        normal_scores = [s for s in violence_scores if s["label"] == "normal"]
        
        print(f"Total inference runs: {len(violence_scores)}")
        if violence_detected:
            print(f"🔴 VIOLENCE DETECTED in {len(violence_detected)} frames:")
            for v in violence_detected:
                print(f"   Frame {v['frame']:04d} ({v['time']:.1f}s): {v['label']} at {v['conf']*100:.1f}%")
        else:
            print(f"🟢 No violence detected. All {len(normal_scores)} checks returned 'normal'")
            if normal_scores:
                normal_confs = [s["conf"] for s in normal_scores]
                print(f"   Normal confidence: min={min(normal_confs)*100:.1f}%, max={max(normal_confs)*100:.1f}%, avg={np.mean(normal_confs)*100:.1f}%")
    else:
        print("⚠️ VideoMAE never ran (need 16 frames + people in scene)")
    
    print(f"\n--- Running Detection (MediaPipe) ---")
    if running_detections:
        running_found = [r for r in running_detections if r["behavior"] == "running"]
        walking = [r for r in running_detections if r["behavior"] == "walking/standing"]
        unknown = [r for r in running_detections if r["behavior"] == "unknown"]
        
        print(f"Total person-frame analyses: {len(running_detections)}")
        if running_found:
            print(f"🟠 RUNNING DETECTED in {len(running_found)} person-frames:")
            for r in running_found[:10]:
                print(f"   Frame {r['frame']:04d} ({r['time']:.1f}s): ID{r['track_id']} running at {r['conf']*100:.1f}%")
        else:
            print(f"🟢 No running detected. {len(walking)} walking/standing, {len(unknown)} unknown")
    else:
        print("⚠️ No people detected in any frame")
    
    print(f"\n--- Alerts ---")
    if all_alerts:
        print(f"🚨 {len(all_alerts)} ALERTS TRIGGERED:")
        for a in all_alerts:
            alert = a["alert"]
            print(f"   [{a['time']:.1f}s] [{alert.severity}] {alert.threat_type} "
                  f"(confidence: {alert.confidence*100:.1f}%)")
    else:
        print("No alerts triggered")
    
    print(f"\n--- Output ---")
    print(f"Annotated video saved to: {os.path.abspath(output_path)}")
    print("=" * 70)


if __name__ == "__main__":
    main()