Lower confidence threshold to 0.001 and add detailed confidence analysis

config.yaml

@@ -1,6 +1,6 @@
 model:
   path: 'VietCat/GTSRB-Model/models/GTSRB.pt' # Path to the YOLO model on Hugging Face Hub (will be downloaded automatically)
-  confidence_threshold: 0.15 # Minimum confidence for detections
+  confidence_threshold: 0.001 # Minimum confidence for detections (very low for testing)
 
 inference:
   box_color: (128, 0, 128) # Purple color for bounding boxes (BGR format)

model.py

@@ -196,10 +196,16 @@ class TrafficSignDetector:
         
         # Run with conf=0.0 to get raw predictions (before filtering)
         results_raw = self.model(image, conf=0.0, imgsz=640, iou=0.45)
-        print(f"  - Raw detections (conf=0.0): {len(results_raw[0].boxes) if results_raw else 0}")
+        raw_box_count = len(results_raw[0].boxes) if results_raw else 0
+        print(f"  - Raw detections (conf=0.0): {raw_box_count}")
+        
         if results_raw and len(results_raw[0].boxes) > 0:
-            raw_confs = [float(box.conf[0]) for box in results_raw[0].boxes[:5]]  # First 5
-            print(f"  - Top 5 raw confidences: {[f'{c:.4f}' for c in raw_confs]}")
+            all_raw_confs = [float(box.conf[0]) for box in results_raw[0].boxes]
+            print(f"  - Top 5 raw confidences: {[f'{c:.6f}' for c in sorted(all_raw_confs, reverse=True)[:5]]}")
+            print(f"  - Confidence stats: min={min(all_raw_confs):.6f}, max={max(all_raw_confs):.6f}, mean={np.mean(all_raw_confs):.6f}")
+            print(f"  - Confidences > 0.01: {sum(1 for c in all_raw_confs if c > 0.01)}")
+            print(f"  - Confidences > 0.001: {sum(1 for c in all_raw_confs if c > 0.001)}")
+            print(f"  - Confidences > 0.0001: {sum(1 for c in all_raw_confs if c > 0.0001)}")
         
         # Now run with actual threshold
         results = self.model(image, conf=self.conf_threshold, imgsz=640, iou=0.45)
@@ -256,15 +262,29 @@ class TrafficSignDetector:
         print(f"{'='*80}")
         
         # Analysis and recommendations
+        print(f"\n📋 ANALYSIS & RECOMMENDATIONS:")
+        
+        # Check for raw detections issue
+        if raw_box_count > 0 and max([float(box.conf[0]) for box in results_raw[0].boxes]) < 0.01:
+            print(f"  ⚠️  MODEL CONFIDENCE ISSUE:")
+            print(f"     - Model detects {raw_box_count} objects but all with confidence < 0.01")
+            print(f"     - This indicates the model may not be well-trained for this domain")
+            print(f"     - Possible causes:")
+            print(f"       a) Model trained on different dataset/resolution")
+            print(f"       b) Model underfitted (needs more training epochs)")
+            print(f"       c) Training data does not match inference data")
+            print(f"       d) Model weights not properly saved")
+            print(f"     - Solutions:")
+            print(f"       1) Retrain model with proper hyperparameters (100+ epochs)")
+            print(f"       2) Use augmentation during training")
+            print(f"       3) Check training/validation accuracy was good")
+            print(f"       4) Ensure training data matches inference image types")
+        
         if scale < 0.5:
-            print(f"\n📋 ANALYSIS & RECOMMENDATIONS:")
-            print(f"  1. Objects are too small after resizing (scale={scale:.2f})")
-            print(f"  2. Possible solutions:")
-            print(f"     a) Use larger target size (imgsz=1024 or 1280)")
-            print(f"     b) Use multi-scale detection (detect multiple regions)")
-            print(f"     c) Reduce input image size to avoid extreme scaling")
-            print(f"     d) Check if model was trained with your image dimensions")
-            print(f"  3. Current: {original_image.shape} → {image.shape} (scale {scale:.2f})")
+            print(f"\n  ⚠️  SCALING ISSUE:")
+            print(f"     - Objects too small after resizing (scale={scale:.2f})")
+            print(f"     - Current: {original_image.shape} → {image.shape}")
+            print(f"     - Solutions: use larger imgsz (1024/1280) or smaller input images")
         
         print()
         return original_image, preprocessed_display