#!/usr/bin/env python3 """ Script de test automatisé du modèle de détection Permet de vérifier le fonctionnement et les performances """ import cv2 import numpy as np from pathlib import Path import argparse import json from ultralytics import YOLO import time from tqdm import tqdm class ModelTester: def __init__(self, model_path, test_images_dir): self.model_path = Path(model_path) self.test_images_dir = Path(test_images_dir) self.model = None self.results = { 'total_images': 0, 'total_detections': 0, 'avg_inference_time': 0, 'avg_confidence': 0, 'images_with_detections': 0, 'per_image_results': [] } def load_model(self): """Charge le modèle YOLO""" print(f"📦 Chargement du modèle: {self.model_path}") self.model = YOLO(str(self.model_path)) print("✅ Modèle chargé") def test_single_image(self, image_path, visualize=False, save_path=None): """Teste sur une seule image""" img = cv2.imread(str(image_path)) if img is None: print(f"❌ Impossible de charger: {image_path}") return None # Inférence start_time = time.time() results = self.model(img, verbose=False)[0] inference_time = (time.time() - start_time) * 1000 # ms # Extraire les détections detections = [] for box in results.boxes: x1, y1, x2, y2 = box.xyxy[0].cpu().numpy() conf = float(box.conf[0]) cls = int(box.cls[0]) detections.append({ 'bbox': [float(x1), float(y1), float(x2), float(y2)], 'confidence': conf, 'class': cls, 'class_name': self.model.names[cls] }) result = { 'image': image_path.name, 'inference_time_ms': inference_time, 'num_detections': len(detections), 'detections': detections } # Visualisation if visualize or save_path: vis_img = self.visualize_detections(img, detections) if visualize: cv2.imshow('Détections', vis_img) cv2.waitKey(0) if save_path: cv2.imwrite(str(save_path), vis_img) return result def visualize_detections(self, img, detections): """Dessine les détections sur l'image""" vis_img = img.copy() for det in detections: x1, y1, x2, y2 = [int(v) for v in det['bbox']] conf = det['confidence'] class_name = det['class_name'] # Couleur selon la classe color = (0, 255, 0) if class_name == 'queen' else (255, 0, 0) # Rectangle cv2.rectangle(vis_img, (x1, y1), (x2, y2), color, 2) # Label label = f"{class_name}: {conf:.2f}" (w, h), _ = cv2.getTextSize(label, cv2.FONT_HERSHEY_SIMPLEX, 0.6, 2) cv2.rectangle(vis_img, (x1, y1-h-10), (x1+w, y1), color, -1) cv2.putText(vis_img, label, (x1, y1-5), cv2.FONT_HERSHEY_SIMPLEX, 0.6, (255, 255, 255), 2) return vis_img def test_directory(self, visualize=False, save_results=True): """Teste sur un dossier d'images""" print(f"\n🧪 Test sur le dossier: {self.test_images_dir}") # Trouver toutes les images image_files = list(self.test_images_dir.glob('*.jpg')) + \ list(self.test_images_dir.glob('*.jpeg')) + \ list(self.test_images_dir.glob('*.png')) if not image_files: print("❌ Aucune image trouvée") return print(f"📊 {len(image_files)} images à tester") # Dossier de sortie pour visualisations output_dir = Path('test_results') output_dir.mkdir(exist_ok=True) # Tester chaque image inference_times = [] confidences = [] for img_path in tqdm(image_files, desc="Test en cours"): save_path = output_dir / f"result_{img_path.name}" if save_results else None result = self.test_single_image(img_path, visualize=visualize, save_path=save_path) if result: self.results['per_image_results'].append(result) inference_times.append(result['inference_time_ms']) if result['num_detections'] > 0: self.results['images_with_detections'] += 1 self.results['total_detections'] += result['num_detections'] for det in result['detections']: confidences.append(det['confidence']) # Calculer les statistiques self.results['total_images'] = len(image_files) if inference_times: self.results['avg_inference_time'] = np.mean(inference_times) self.results['min_inference_time'] = np.min(inference_times) self.results['max_inference_time'] = np.max(inference_times) if confidences: self.results['avg_confidence'] = np.mean(confidences) self.results['min_confidence'] = np.min(confidences) self.results['max_confidence'] = np.max(confidences) # Afficher le rapport self.print_report() # Sauvegarder JSON if save_results: with open(output_dir / 'test_results.json', 'w') as f: json.dump(self.results, f, indent=2) print(f"\n💾 Résultats sauvegardés dans: {output_dir}") def print_report(self): """Affiche le rapport de test""" print("\n" + "="*60) print("📋 RAPPORT DE TEST") print("="*60) print(f"\n📊 Statistiques globales:") print(f" Images testées: {self.results['total_images']}") print(f" Images avec détections: {self.results['images_with_detections']} " f"({100*self.results['images_with_detections']/max(1,self.results['total_images']):.1f}%)") print(f" Détections totales: {self.results['total_detections']}") if self.results['total_detections'] > 0: avg_per_image = self.results['total_detections'] / self.results['images_with_detections'] print(f" Détections par image (avg): {avg_per_image:.2f}") print(f"\n⏱️ Performance:") print(f" Temps d'inférence moyen: {self.results.get('avg_inference_time', 0):.2f} ms") print(f" Min: {self.results.get('min_inference_time', 0):.2f} ms") print(f" Max: {self.results.get('max_inference_time', 0):.2f} ms") if 'avg_inference_time' in self.results: fps = 1000 / self.results['avg_inference_time'] print(f" FPS théorique: {fps:.1f}") print(f"\n🎯 Confiance:") if 'avg_confidence' in self.results: print(f" Confiance moyenne: {self.results['avg_confidence']:.3f}") print(f" Min: {self.results['min_confidence']:.3f}") print(f" Max: {self.results['max_confidence']:.3f}") else: print(" Aucune détection") # Distribution par classe class_counts = {} for img_result in self.results['per_image_results']: for det in img_result['detections']: class_name = det['class_name'] class_counts[class_name] = class_counts.get(class_name, 0) + 1 if class_counts: print(f"\n📈 Distribution par classe:") for class_name, count in sorted(class_counts.items(), key=lambda x: -x[1]): pct = 100 * count / self.results['total_detections'] print(f" {class_name}: {count} ({pct:.1f}%)") # Évaluation print(f"\n💡 Évaluation:") if self.results.get('avg_inference_time', 999) < 100: print(" ✅ Performance excellente (<100ms)") elif self.results.get('avg_inference_time', 999) < 150: print(" ✅ Performance bonne (<150ms)") else: print(" ⚠️ Performance à améliorer (>150ms)") detection_rate = self.results['images_with_detections'] / max(1, self.results['total_images']) if detection_rate > 0.7: print(" ✅ Taux de détection élevé (>70%)") elif detection_rate > 0.5: print(" ⚠️ Taux de détection moyen (50-70%)") else: print(" ❌ Taux de détection faible (<50%)") if self.results.get('avg_confidence', 0) > 0.7: print(" ✅ Confiance élevée (>0.7)") elif self.results.get('avg_confidence', 0) > 0.5: print(" ⚠️ Confiance moyenne (0.5-0.7)") else: print(" ❌ Confiance faible (<0.5)") def main(): parser = argparse.ArgumentParser( description='Test automatisé du modèle de détection' ) parser.add_argument('--model', type=str, required=True, help='Chemin vers le modèle (.pt)') parser.add_argument('--images', type=str, required=True, help='Dossier contenant les images de test') parser.add_argument('--visualize', action='store_true', help='Afficher les résultats image par image') parser.add_argument('--no-save', action='store_true', help='Ne pas sauvegarder les résultats') args = parser.parse_args() tester = ModelTester(args.model, args.images) tester.load_model() tester.test_directory( visualize=args.visualize, save_results=not args.no_save ) if args.visualize: cv2.destroyAllWindows() if __name__ == '__main__': main()