File size: 40,346 Bytes
ac9ec2c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
import cv2
import numpy as np
from ultralytics import YOLO
import torch
import time
from datetime import datetime
import os
import json
from threading import Thread
import queue
from typing import Dict, List, Tuple, Optional
import requests

class SafetyDetector:
    """
    Real-time safety compliance detection system using YOLO for object detection.
    Detects people and safety equipment like hard hats, safety vests, and safety glasses.
    """
    
    def __init__(self, model_path: Optional[str] = None, confidence_threshold: float = 0.5):
        """
        Initialize the safety detector with a specialized PPE detection model.
        
        Args:
            model_path: Path to custom model, if None will download PPE model
            confidence_threshold: Minimum confidence for detections
        """
        self.confidence_threshold = confidence_threshold
        self.device = 'cuda' if torch.cuda.is_available() else 'cpu'
        
        # Stricter confidence thresholds for different equipment types to reduce false positives
        self.equipment_confidence_thresholds = {
            'hardhat': 0.7,      # Higher threshold for hard hats (hair confusion)
            'safety_vest': 0.75,  # Higher threshold for safety vests (clothing confusion)
            'mask': 0.6,         # Moderate threshold for masks
            'person': 0.5,       # Standard threshold for people
            'no_hardhat': 0.6,   # Moderate threshold for NO- detections
            'no_safety_vest': 0.6,
            'no_mask': 0.6
        }
        
        # Try to load a specialized PPE detection model
        self.model = self._load_ppe_model(model_path)
        
        # PPE class names - these are the actual classes we expect from PPE models
        self.ppe_classes = {
            'hardhat': ['Hardhat', 'hardhat', 'helmet', 'hard hat'],
            'safety_vest': ['Safety Vest', 'safety vest', 'vest', 'safety-vest', 'Safety-Vest'],
            'no_hardhat': ['NO-Hardhat', 'no-hardhat', 'no hardhat', 'NO-Helmet'],
            'no_safety_vest': ['NO-Safety Vest', 'no-safety-vest', 'no safety vest', 'NO-Safety-Vest'],
            'person': ['Person', 'person'],
            'mask': ['Mask', 'mask'],
            'no_mask': ['NO-Mask', 'no-mask', 'no mask'],
            'safety_gloves': ['Safety Gloves', 'safety-gloves', 'gloves', 'Gloves'],
            'safety_glasses': ['Safety Glasses', 'safety-glasses', 'glasses', 'Safety-Glasses'],
            'hearing_protection': ['Hearing Protection', 'hearing-protection', 'ear protection']
        }
        
        print(f"Using device: {self.device}")
        print(f"Loaded PPE detection model with stricter confidence thresholds")
        print(f"Equipment thresholds: {self.equipment_confidence_thresholds}")
        
        # Colors for bounding boxes
        self.colors = {
            'person': (0, 255, 0),          # Green for compliant person
            'violation': (0, 0, 255),       # Red for safety violation
            'equipment': (255, 255, 0),     # Yellow for safety equipment
            'warning': (0, 165, 255)        # Orange for warnings
        }
        
        # Violation tracking
        self.violations = []
        self.violation_images_dir = "violation_captures"
        os.makedirs(self.violation_images_dir, exist_ok=True)
        
    def _load_ppe_model(self, model_path: Optional[str] = None) -> YOLO:
        """Load a specialized PPE detection model."""
        if model_path and os.path.exists(model_path):
            print(f"Loading custom model from {model_path}")
            return YOLO(model_path)
        
        # Try to download YOLOv8-compatible PPE models
        ppe_model_urls = [
            # Try the snehilsanyal YOLOv8 PPE model (best.pt)
            "https://github.com/snehilsanyal/Construction-Site-Safety-PPE-Detection/raw/main/models/best.pt",
            # Try mayank13-01 YOLOv8 PPE model
            "https://github.com/mayank13-01/Yolov8-PPE/raw/main/YOLO-Weights/ppe.pt"
        ]
        
        for i, url in enumerate(ppe_model_urls):
            try:
                model_filename = f"ppe_yolov8_model_{i}.pt"
                if not os.path.exists(model_filename):
                    print(f"Downloading PPE detection model from {url}...")
                    response = requests.get(url, timeout=60)
                    if response.status_code == 200:
                        with open(model_filename, 'wb') as f:
                            f.write(response.content)
                        print(f"Downloaded PPE model successfully as {model_filename}")
                
                if os.path.exists(model_filename):
                    print(f"Loading YOLOv8 PPE model from {model_filename}")
                    model = YOLO(model_filename)
                    
                    # Test if the model loads properly
                    classes = self._get_model_classes(model)
                    print(f"Model classes: {classes}")
                    
                    # Check if it has PPE-related classes
                    ppe_related = any(
                        any(keyword in str(cls).lower() for keyword in ['hardhat', 'vest', 'helmet', 'mask', 'person'])
                        for cls in classes
                    )
                    
                    if ppe_related:
                        print(f"✅ Found PPE-capable model with {len(classes)} classes")
                        return model
                    else:
                        print(f"⚠️  Model doesn't seem to have PPE classes: {classes}")
                        
            except Exception as e:
                print(f"Failed to download/load from {url}: {e}")
                continue
        
        # Fallback to YOLOv8 with a warning
        print("⚠️  Warning: Could not load specialized PPE model, falling back to YOLOv8n")
        print("   Note: YOLOv8n can detect people but not safety equipment")
        return YOLO('yolov8n.pt')
    
    def _get_model_classes(self, model=None) -> List[str]:
        """Get the list of classes the model can detect."""
        if model is None:
            model = self.model
        if hasattr(model, 'names'):
            return list(model.names.values())
        return []
    
    def _get_class_category(self, class_name: str) -> str:
        """Map detected class name to our safety categories."""
        class_name_lower = class_name.lower()
        
        for category, variations in self.ppe_classes.items():
            for variation in variations:
                if variation.lower() in class_name_lower or class_name_lower in variation.lower():
                    return category
        
        return class_name_lower
    
    def detect_safety_violations(self, frame: np.ndarray) -> Dict:
        """
        Detect safety violations in the given frame with improved accuracy.
        
        Returns:
            Dictionary containing detection results and violations
        """
        start_time = time.time()
        
        # Run detection with optimized settings for speed
        results = self.model(frame, conf=0.3, verbose=False, imgsz=640, half=False)
        
        detections = []
        people_count = 0
        safety_equipment_detected = {
            'hardhat': 0,
            'safety_vest': 0,
            'safety_gloves': 0,
            'safety_glasses': 0,
            'hearing_protection': 0,
            'mask': 0
        }
        violations = []
        no_equipment_detections = []  # Track NO- detections separately
        
        # Process detections with stricter filtering
        for r in results:
            boxes = r.boxes
            if boxes is not None:
                for box in boxes:
                    # Get detection info
                    x1, y1, x2, y2 = box.xyxy[0].cpu().numpy()
                    confidence = box.conf[0].cpu().numpy()
                    class_id = int(box.cls[0].cpu().numpy())
                    
                    # Get class name
                    if hasattr(self.model, 'names'):
                        class_name = self.model.names[class_id]
                    else:
                        class_name = f"class_{class_id}"
                    
                    # Map to our categories
                    category = self._get_class_category(class_name)
                    
                    # Apply stricter confidence thresholds based on equipment type
                    required_confidence = self.equipment_confidence_thresholds.get(category, self.confidence_threshold)
                    
                    # Skip detections that don't meet the stricter threshold
                    if confidence < required_confidence:
                        continue
                    
                    detection = {
                        'bbox': [int(x1), int(y1), int(x2), int(y2)],
                        'confidence': float(confidence),
                        'class': class_name,
                        'category': category
                    }
                    detections.append(detection)
                    
                    # Count people and safety equipment
                    if category == 'person':
                        people_count += 1
                    elif category in safety_equipment_detected:
                        safety_equipment_detected[category] += 1
                    elif category in ['hardhat', 'safety_vest', 'mask'] and not category.startswith('no_'):
                        safety_equipment_detected[category] += 1
                    
                    # Handle negative detections (NO-Hardhat, NO-Mask, etc.)
                    # These indicate violations - a person without required equipment
                    if category.startswith('no_'):
                        equipment_type = category.replace('no_', '')
                        if equipment_type in ['hardhat', 'safety_vest', 'mask']:
                            no_equipment_detections.append({
                                'type': f'missing_{equipment_type}',
                                'severity': 'high',
                                'description': f'Person detected without {equipment_type.replace("_", " ").title()}',
                                'bbox': [int(x1), int(y1), int(x2), int(y2)],
                                'confidence': float(confidence),
                                'equipment_type': equipment_type
                            })
        
        # Create violations based on NO- detections (these are more reliable)
        violations.extend(no_equipment_detections)
        
        # If we have people but no NO- detections, check equipment ratios
        if people_count > 0 and len(no_equipment_detections) == 0:
            required_equipment = ['hardhat', 'safety_vest', 'mask']
            
            for equipment in required_equipment:
                detected_count = safety_equipment_detected[equipment]
                
                # If significantly fewer equipment than people, assume violations
                if detected_count < people_count * 0.8:  # Allow some tolerance
                    missing_count = people_count - detected_count
                    equipment_name = equipment.replace("_", " ").title()
                    violations.append({
                        'type': f'missing_{equipment}',
                        'severity': 'high',
                        'description': f'{missing_count} person(s) likely missing {equipment_name}',
                        'count': missing_count
                    })
        
        # Special handling for masks - they're often not detected well
        mask_detected = safety_equipment_detected['mask']
        no_mask_detected = len([v for v in no_equipment_detections if v['equipment_type'] == 'mask'])
        
        if people_count > 0 and mask_detected == 0 and no_mask_detected == 0:
            # No mask detections at all - assume people are not wearing masks
            violations.append({
                'type': 'missing_mask',
                'severity': 'high',
                'description': f'{people_count} person(s) not wearing Face Mask',
                'count': people_count
            })
        
        processing_time = time.time() - start_time
        
        return {
            'detections': detections,
            'people_count': people_count,
            'safety_equipment': safety_equipment_detected,
            'violations': violations,
            'processing_time': processing_time,
            'fps': 1.0 / processing_time if processing_time > 0 else 0
        }
    
    def draw_detections(self, frame: np.ndarray, results: Dict) -> np.ndarray:
        """
        Draw premium bounding boxes only for POSITIVE equipment detections.
        No boxes for missing equipment - violations shown through person status only.
        
        Args:
            frame: Input frame
            results: Detection results containing detections, violations, etc.
            
        Returns:
            Annotated frame with premium styling
        """
        annotated_frame = frame.copy()
        height, width = annotated_frame.shape[:2]
        
        # Create overlay for semi-transparent effects
        overlay = annotated_frame.copy()
        
        # Premium color scheme
        colors = {
            'person_compliant': (46, 204, 113),      # Emerald green
            'person_violation': (231, 76, 60),       # Red
            'equipment': (52, 152, 219),             # Blue
            'hardhat': (46, 204, 113),               # Green
            'safety_vest': (241, 196, 15),           # Yellow
            'mask': (0, 191, 255),                   # Deep sky blue
            'violation_bg': (231, 76, 60),           # Red background
            'text_bg': (44, 62, 80),                 # Dark blue-gray
            'text_primary': (255, 255, 255),         # White
            'text_secondary': (149, 165, 166),       # Light gray
            'shadow': (0, 0, 0),                     # Black shadow
            'accent': (155, 89, 182),                # Purple accent
        }
        
        # Track people and their compliance status
        people_status = {}
        
        # First pass: categorize people
        for detection in results.get('detections', []):
            class_name = detection['class'].lower()
            bbox = detection['bbox']
            confidence = detection['confidence']
            
            if 'person' in class_name:
                person_id = f"person_{bbox[0]}_{bbox[1]}"
                people_status[person_id] = {
                    'bbox': bbox,
                    'confidence': confidence,
                    'violations': [],
                    'equipment': []
                }
        
        # Map violations to people
        for violation in results.get('violations', []):
            if 'bbox' in violation:
                # This is a specific violation with a bounding box (from NO- detections)
                violation_bbox = violation['bbox']
                # Find the closest person to this violation
                closest_person = None
                min_distance = float('inf')
                
                for person_id, person_data in people_status.items():
                    person_bbox = person_data['bbox']
                    # Calculate distance between violation and person
                    distance = abs(violation_bbox[0] - person_bbox[0]) + abs(violation_bbox[1] - person_bbox[1])
                    if distance < min_distance:
                        min_distance = distance
                        closest_person = person_id
                
                if closest_person and min_distance < 100:  # Within reasonable distance
                    violation_type = violation['type'].replace('missing_', '')
                    people_status[closest_person]['violations'].append(violation_type)
            else:
                # General violation - apply to all people (when equipment count < people count)
                violation_type = violation['type'].replace('missing_', '')
                for person_id in people_status:
                    people_status[person_id]['violations'].append(violation_type)
        
        # If no specific violations detected but people are present, assume they're missing all required equipment
        if len(people_status) > 0 and len(results.get('violations', [])) == 0:
            # Check if we have any positive equipment detections
            equipment_detected = any(
                detection['category'] in ['hardhat', 'safety_vest', 'mask'] 
                for detection in results.get('detections', [])
                if detection['category'] in ['hardhat', 'safety_vest', 'mask']
            )
            
            # If no equipment detected at all, mark all people as having violations
            if not equipment_detected:
                for person_id in people_status:
                    people_status[person_id]['violations'] = ['hardhat', 'safety_vest', 'mask']
        
        # ONLY draw POSITIVE equipment detections (when equipment IS being worn)
        for detection in results.get('detections', []):
            class_name = detection['class'].lower()
            category = detection.get('category', '')
            
            # Skip people and NO- detections - we only want positive equipment
            if 'person' in class_name or 'no-' in class_name or 'no_' in category:
                continue
            
            # Only draw positive equipment detections
            if category in ['hardhat', 'safety_vest', 'mask'] or any(equip in class_name for equip in ['hardhat', 'vest', 'helmet', 'safety', 'mask']):
                bbox = detection['bbox']
                confidence = detection['confidence']
                
                # Choose color and label based on equipment type
                if any(x in class_name for x in ['hardhat', 'helmet']) or category == 'hardhat':
                    color = colors['hardhat']
                    equipment_type = "Hard Hat ✓"
                elif 'vest' in class_name or category == 'safety_vest':
                    color = colors['safety_vest']
                    equipment_type = "Safety Vest ✓"
                elif 'mask' in class_name or category == 'mask':
                    color = colors['mask']
                    equipment_type = "Face Mask ✓"
                else:
                    color = colors['equipment']
                    equipment_type = "Safety Equipment ✓"
                
                # Draw equipment with premium styling
                self._draw_premium_bbox(overlay, annotated_frame, bbox, color, 
                                      equipment_type, confidence, 
                                      bbox_type="equipment", colors=colors)
        
        # Draw people with compliance status (no violation indicators on person boxes)
        for person_id, person_data in people_status.items():
            bbox = person_data['bbox']
            confidence = person_data['confidence']
            violations = person_data['violations']
            
            # Determine person status
            is_compliant = len(violations) == 0
            color = colors['person_compliant'] if is_compliant else colors['person_violation']
            status_text = "COMPLIANT" if is_compliant else "VIOLATION"
            
            # Draw person with premium styling (no violation details on the box)
            self._draw_premium_bbox(overlay, annotated_frame, bbox, color,
                                  f"Person - {status_text}", confidence,
                                  bbox_type="person", violations=None,  # Don't show violation details on person box
                                  colors=colors)
        
        # Blend overlay with original frame for semi-transparent effects
        alpha = 0.15
        cv2.addWeighted(overlay, alpha, annotated_frame, 1 - alpha, 0, annotated_frame)
        
        # Statistics are now handled by the web UI, no overlay needed on video feed
        
        return annotated_frame
    
    def _draw_premium_bbox(self, overlay, frame, bbox, color, label, confidence, 
                          bbox_type="default", violations=None, colors=None):
        """Draw a premium-styled bounding box with advanced visual effects."""
        x1, y1, x2, y2 = map(int, bbox)
        
        # Box dimensions
        box_width = x2 - x1
        box_height = y2 - y1
        
        # Draw shadow first (slightly offset)
        shadow_offset = 3
        shadow_color = colors['shadow']
        cv2.rectangle(overlay, 
                     (x1 + shadow_offset, y1 + shadow_offset), 
                     (x2 + shadow_offset, y2 + shadow_offset), 
                     shadow_color, 2)
        
        # Main bounding box with thinner lines
        box_thickness = 2 if bbox_type == "person" else 1
        
        # Draw main rectangle
        cv2.rectangle(frame, (x1, y1), (x2, y2), color, box_thickness)
        
        # Draw corner accents for premium look
        corner_length = min(20, box_width // 4, box_height // 4)
        accent_thickness = box_thickness
        
        # Top-left corner
        cv2.line(frame, (x1, y1), (x1 + corner_length, y1), color, accent_thickness)
        cv2.line(frame, (x1, y1), (x1, y1 + corner_length), color, accent_thickness)
        
        # Top-right corner
        cv2.line(frame, (x2, y1), (x2 - corner_length, y1), color, accent_thickness)
        cv2.line(frame, (x2, y1), (x2, y1 + corner_length), color, accent_thickness)
        
        # Bottom-left corner
        cv2.line(frame, (x1, y2), (x1 + corner_length, y2), color, accent_thickness)
        cv2.line(frame, (x1, y2), (x1, y2 - corner_length), color, accent_thickness)
        
        # Bottom-right corner
        cv2.line(frame, (x2, y2), (x2 - corner_length, y2), color, accent_thickness)
        cv2.line(frame, (x2, y2), (x2, y2 - corner_length), color, accent_thickness)
        
        # Prepare label text
        confidence_text = f"{confidence:.1%}"
        main_text = f"{label}"
        
        # Calculate text dimensions
        font = cv2.FONT_HERSHEY_SIMPLEX
        font_scale = 0.5
        thickness = 1
        
        (main_w, main_h), _ = cv2.getTextSize(main_text, font, font_scale, thickness)
        (conf_w, conf_h), _ = cv2.getTextSize(confidence_text, font, font_scale - 0.1, thickness - 1)
        
        # Label background dimensions
        label_height = max(main_h, conf_h) + 12
        label_width = max(main_w, conf_w) + 16
        
        # Position label (above box if space available, otherwise below)
        if y1 - label_height - 5 > 0:
            label_y = y1 - label_height - 5
        else:
            label_y = y2 + 5
        
        label_x = x1
        
        # Ensure label stays within frame
        if label_x + label_width > frame.shape[1]:
            label_x = frame.shape[1] - label_width - 5
        if label_x < 0:
            label_x = 5
        
        # Draw label background with gradient effect
        bg_color = colors['text_bg']
        
        # Main background
        cv2.rectangle(overlay, 
                     (label_x, label_y), 
                     (label_x + label_width, label_y + label_height), 
                     bg_color, -1)
        
        # Colored top border
        cv2.rectangle(frame, 
                     (label_x, label_y), 
                     (label_x + label_width, label_y + 4), 
                     color, -1)
        
        # Add subtle border
        cv2.rectangle(frame, 
                     (label_x, label_y), 
                     (label_x + label_width, label_y + label_height), 
                     color, 1)
        
        # Draw main text
        text_y = label_y + main_h + 6
        cv2.putText(frame, main_text, 
                   (label_x + 8, text_y), 
                   font, font_scale, colors['text_primary'], thickness)
        
        # Draw confidence text
        conf_y = text_y + conf_h + 4
        cv2.putText(frame, confidence_text, 
                   (label_x + 8, conf_y), 
                   font, font_scale - 0.1, colors['text_secondary'], max(1, thickness - 1))
        
        # Draw violation indicators for people (only if violations are provided)
        if bbox_type == "person" and violations is not None and len(violations) > 0:
            self._draw_violation_indicators(frame, overlay, x1, y1, x2, y2, violations, colors)
    
    def _draw_violation_indicators(self, frame, overlay, x1, y1, x2, y2, violations, colors):
        """Draw violation indicators with premium styling."""
        # Warning icon position (top-right of bounding box)
        icon_size = 24
        icon_x = x2 - icon_size - 5
        icon_y = y1 + 5
        
        # Draw warning background circle
        cv2.circle(overlay, (icon_x + icon_size//2, icon_y + icon_size//2), 
                  icon_size//2, colors['violation_bg'], -1)
        cv2.circle(frame, (icon_x + icon_size//2, icon_y + icon_size//2), 
                  icon_size//2, colors['violation_bg'], 2)
        
        # Draw exclamation mark
        center_x = icon_x + icon_size//2
        center_y = icon_y + icon_size//2
        
        # Exclamation line
        cv2.line(frame, (center_x, center_y - 6), (center_x, center_y + 2), 
                colors['text_primary'], 2)
        # Exclamation dot
        cv2.circle(frame, (center_x, center_y + 5), 1, colors['text_primary'], -1)
        
        # Draw violation list below the person if space allows
        violation_text = "Missing: " + ", ".join(violations)
        font = cv2.FONT_HERSHEY_SIMPLEX
        font_scale = 0.5
        thickness = 1
        
        (text_w, text_h), _ = cv2.getTextSize(violation_text, font, font_scale, thickness)
        
        # Position violation text
        viol_x = x1
        viol_y = y2 + text_h + 8
        
        # Ensure text stays within frame
        if viol_y + text_h > frame.shape[0]:
            viol_y = y1 - text_h - 8
        if viol_x + text_w > frame.shape[1]:
            viol_x = frame.shape[1] - text_w - 5
        
        # Draw violation text background
        padding = 4
        cv2.rectangle(overlay, 
                     (viol_x - padding, viol_y - text_h - padding), 
                     (viol_x + text_w + padding, viol_y + padding), 
                     colors['violation_bg'], -1)
        
        # Draw violation text
        cv2.putText(frame, violation_text, 
                   (viol_x, viol_y), 
                   font, font_scale, colors['text_primary'], thickness)
    
    def _draw_statistics_overlay(self, frame, results, colors, width, height):
        """Draw statistics overlay with premium styling."""
        # Statistics data
        people_count = results.get('people_count', 0)
        violations = results.get('violations', [])
        violation_count = len(violations)
        compliant_count = people_count - violation_count
        compliance_rate = (compliant_count / max(people_count, 1)) * 100
        
        # Statistics text
        stats = [
            f"People: {people_count}",
            f"Compliant: {compliant_count}",
            f"Violations: {violation_count}",
            f"Compliance: {compliance_rate:.1f}%"
        ]
        
        # Text properties
        font = cv2.FONT_HERSHEY_SIMPLEX
        font_scale = 0.7
        thickness = 2
        
        # Calculate background size
        max_text_width = 0
        total_height = 0
        line_heights = []
        
        for text in stats:
            (text_w, text_h), _ = cv2.getTextSize(text, font, font_scale, thickness)
            max_text_width = max(max_text_width, text_w)
            line_heights.append(text_h)
            total_height += text_h + 8
        
        # Background dimensions
        bg_width = max_text_width + 24
        bg_height = total_height + 16
        
        # Position (top-left corner)
        bg_x = 20
        bg_y = 20
        
        # Draw semi-transparent background
        overlay = frame.copy()
        cv2.rectangle(overlay, 
                     (bg_x, bg_y), 
                     (bg_x + bg_width, bg_y + bg_height), 
                     colors['text_bg'], -1)
        cv2.addWeighted(overlay, 0.8, frame, 0.2, 0, frame)
        
        # Draw border
        cv2.rectangle(frame, 
                     (bg_x, bg_y), 
                     (bg_x + bg_width, bg_y + bg_height), 
                     colors['accent'], 2)
        
        # Draw statistics text
        current_y = bg_y + 24
        for i, text in enumerate(stats):
            # Choose color based on statistic type
            if "Violations:" in text and violation_count > 0:
                text_color = colors['person_violation']
            elif "Compliant:" in text:
                text_color = colors['person_compliant']
            elif "Compliance:" in text:
                if compliance_rate >= 80:
                    text_color = colors['person_compliant']
                elif compliance_rate >= 60:
                    text_color = colors['safety_vest']
                else:
                    text_color = colors['person_violation']
            else:
                text_color = colors['text_primary']
            
            cv2.putText(frame, text, 
                       (bg_x + 12, current_y), 
                       font, font_scale, text_color, thickness)
            current_y += line_heights[i] + 8
    
    def get_model_classes(self) -> List[str]:
        """Get the list of classes the model can detect."""
        return self._get_model_classes()
    
    def test_detection(self, test_image_path: str = None):
        """Test the detector with a sample image or webcam."""
        if test_image_path and os.path.exists(test_image_path):
            frame = cv2.imread(test_image_path)
            if frame is not None:
                results = self.detect_safety_violations(frame)
                output = self.draw_detections(frame, results)
                
                print(f"Detected classes: {[d['class'] for d in results['detections']]}")
                print(f"Available model classes: {self.get_model_classes()}")
                
                cv2.imshow('PPE Detection Test', output)
                cv2.waitKey(0)
                cv2.destroyAllWindows()
                return results
        else:
            print("Testing with webcam - press 'q' to quit")
            cap = cv2.VideoCapture(0)
            
            while True:
                ret, frame = cap.read()
                if not ret:
                    break
                
                results = self.detect_safety_violations(frame)
                output = self.draw_detections(frame, results)
                
                cv2.imshow('PPE Detection Test', output)
                
                if cv2.waitKey(1) & 0xFF == ord('q'):
                    break
            
            cap.release()
            cv2.destroyAllWindows()

    def analyze_safety_compliance(self, detections: List[Dict]) -> Dict:
        """
        Analyze safety compliance based on detected objects.
        
        Args:
            detections: List of detected objects
            
        Returns:
            Dictionary with compliance analysis
        """
        people_detected = []
        safety_equipment = []
        
        # Separate people and safety equipment
        for detection in detections:
            if detection['class'].lower() == 'person':
                people_detected.append(detection)
            elif any(equipment in detection['class'].lower() 
                    for equipment in ['helmet', 'hardhat', 'vest', 'gloves', 'glasses']):
                safety_equipment.append(detection)
        
        # Analyze compliance for each person
        compliance_results = []
        for person in people_detected:
            person_bbox = person['bbox']
            
            # Check for nearby safety equipment
            nearby_equipment = self._find_nearby_equipment(person_bbox, safety_equipment)
            
            # Determine missing equipment
            required_equipment = ['hardhat', 'safety_vest']
            missing_equipment = []
            
            for equipment in required_equipment:
                if not any(equipment.lower() in item['class'].lower() 
                          for item in nearby_equipment):
                    missing_equipment.append(equipment)
            
            compliance_results.append({
                'person': person,
                'nearby_equipment': nearby_equipment,
                'missing_equipment': missing_equipment,
                'is_compliant': len(missing_equipment) == 0,
                'compliance_score': 1.0 - (len(missing_equipment) / len(required_equipment))
            })
        
        return {
            'total_people': len(people_detected),
            'compliant_people': sum(1 for result in compliance_results if result['is_compliant']),
            'violations': sum(len(result['missing_equipment']) for result in compliance_results),
            'compliance_results': compliance_results,
            'overall_compliance_rate': (
                sum(result['compliance_score'] for result in compliance_results) / 
                max(len(compliance_results), 1)
            )
        }
    
    def _find_nearby_equipment(self, person_bbox: List[int], equipment_list: List[Dict], 
                               proximity_threshold: float = 0.3) -> List[Dict]:
        """Find safety equipment near a person."""
        nearby_equipment = []
        
        person_center_x = (person_bbox[0] + person_bbox[2]) / 2
        person_center_y = (person_bbox[1] + person_bbox[3]) / 2
        
        for equipment in equipment_list:
            equip_bbox = equipment['bbox']
            equip_center_x = (equip_bbox[0] + equip_bbox[2]) / 2
            equip_center_y = (equip_bbox[1] + equip_bbox[3]) / 2
            
            # Calculate normalized distance
            distance = np.sqrt((person_center_x - equip_center_x)**2 + 
                             (person_center_y - equip_center_y)**2)
            
            # Normalize by image diagonal (assuming standard frame size)
            normalized_distance = distance / 1000  # Adjust based on typical frame size
            
            if normalized_distance < proximity_threshold:
                nearby_equipment.append(equipment)
        
        return nearby_equipment
    
    def draw_annotations(self, frame: np.ndarray, analysis: Dict) -> np.ndarray:
        """
        Draw bounding boxes and annotations on the frame.
        
        Args:
            frame: Input frame
            analysis: Safety compliance analysis results
            
        Returns:
            Annotated frame
        """
        annotated_frame = frame.copy()
        
        # Draw safety equipment
        for equipment in analysis['safety_equipment']:
            bbox = equipment['bbox']
            cv2.rectangle(annotated_frame, (bbox[0], bbox[1]), (bbox[2], bbox[3]), 
                         self.colors['equipment'], 2)
            
            label = f"{equipment.get('equipment_type', equipment['class'])}: {equipment['confidence']:.2f}"
            cv2.putText(annotated_frame, label, (bbox[0], bbox[1] - 10), 
                       cv2.FONT_HERSHEY_SIMPLEX, 0.5, self.colors['equipment'], 2)
        
        # Draw people with compliance status
        for result in analysis['compliance_results']:
            person = result['person']
            bbox = person['bbox']
            
            # Choose color based on compliance
            color = self.colors['person'] if result['is_compliant'] else self.colors['violation']
            
            # Draw bounding box
            cv2.rectangle(annotated_frame, (bbox[0], bbox[1]), (bbox[2], bbox[3]), color, 3)
            
            # Create status label
            status = "COMPLIANT" if result['is_compliant'] else "VIOLATION"
            confidence_text = f"Person: {person['confidence']:.2f}"
            
            # Draw labels
            cv2.putText(annotated_frame, status, (bbox[0], bbox[1] - 30), 
                       cv2.FONT_HERSHEY_SIMPLEX, 0.7, color, 2)
            cv2.putText(annotated_frame, confidence_text, (bbox[0], bbox[1] - 10), 
                       cv2.FONT_HERSHEY_SIMPLEX, 0.5, color, 2)
            
            # Show missing equipment
            if result['missing_equipment']:
                missing_text = f"Missing: {', '.join(result['missing_equipment'])}"
                cv2.putText(annotated_frame, missing_text, (bbox[0], bbox[3] + 20), 
                           cv2.FONT_HERSHEY_SIMPLEX, 0.5, self.colors['violation'], 2)
        
        # Draw summary statistics
        summary_text = [
            f"Total People: {analysis['total_people']}",
            f"Compliant: {analysis['compliant_people']}",
            f"Violations: {analysis['violations']}",
            f"Compliance Rate: {(analysis['compliant_people']/max(analysis['total_people'],1)*100):.1f}%"
        ]
        
        for i, text in enumerate(summary_text):
            cv2.putText(annotated_frame, text, (10, 30 + i * 25), 
                       cv2.FONT_HERSHEY_SIMPLEX, 0.6, (255, 255, 255), 2)
        
        return annotated_frame
    
    def capture_violation(self, frame: np.ndarray, violation_data: Dict) -> str:
        """
        Capture and save an image when a safety violation is detected.
        
        Args:
            frame: Current frame
            violation_data: Information about the violation
            
        Returns:
            Path to saved image
        """
        timestamp = datetime.now().strftime("%Y%m%d_%H%M%S_%f")[:-3]
        filename = f"violation_{timestamp}.jpg"
        filepath = os.path.join(self.violation_images_dir, filename)
        
        # Save the frame
        cv2.imwrite(filepath, frame)
        
        # Save violation metadata
        metadata = {
            'timestamp': datetime.now().isoformat(),
            'filename': filename,
            'violation_data': violation_data
        }
        
        metadata_file = filepath.replace('.jpg', '_metadata.json')
        with open(metadata_file, 'w') as f:
            json.dump(metadata, f, indent=2)
        
        self.violations.append(metadata)
        return filepath
    
    def process_frame(self, frame: np.ndarray) -> Tuple[np.ndarray, Dict]:
        """
        Process a single frame for safety monitoring.
        
        Args:
            frame: Input video frame
            
        Returns:
            Tuple of (annotated_frame, analysis_results)
        """
        # Detect objects and get safety violations
        results = self.detect_safety_violations(frame)
        
        # Draw detections on frame using the main drawing method
        annotated_frame = self.draw_detections(frame, results)
        
        return annotated_frame, {
            'detections': results['detections'],
            'people_count': results['people_count'],
            'safety_equipment': results['safety_equipment'],
            'violations': results['violations'],
            'violation_summary': self.get_violation_summary(),
            'frame_stats': {
                'processing_time': results['processing_time'],
                'fps': results['fps'],
                'detection_count': len(results['detections'])
            }
        }
    
    def get_violation_summary(self) -> Dict:
        """Get a summary of recent violations."""
        # This would typically connect to a database or log file
        # For now, return a placeholder
        return {
            'total_violations_today': 0,
            'most_common_violation': 'missing_hardhat',
            'compliance_trend': []  # Could track compliance over time
        }

if __name__ == "__main__":
    # Test the detector
    detector = SafetyDetector()
    print("Available classes:", detector.get_model_classes())
    detector.test_detection()