Update app.py

app.py

@@ -16,18 +16,27 @@ client = Groq(api_key=GROQ_API_KEY)
 
 # Step 1: Load YOLO model
 model = YOLO('yolov8n.pt')  # Ensure the model path is correct
+model.conf = 0.3  # Lower confidence threshold to 30%
 
 # Function to detect weeds, classify types, and analyze severity
 def detect_weeds_and_classify(image):
     results = model(image)  # Run inference on the image
     detections = results[0]
-    image = Image.fromarray(np.uint8(image))
 
-    boxes = detections.boxes.xyxy.cpu().numpy()  # Bounding boxes
-    labels = detections.boxes.cls.cpu().numpy()  # Class labels
-    confidences = detections.boxes.conf.cpu().numpy()  # Confidence scores
+    # Debugging output
+    print("Detection results:", detections)
+
+    boxes = detections.boxes.xyxy.cpu().numpy() if detections.boxes is not None else []
+    labels = detections.boxes.cls.cpu().numpy() if detections.boxes is not None else []
+    confidences = detections.boxes.conf.cpu().numpy() if detections.boxes is not None else []
+
+    if len(boxes) == 0:
+        print("No objects detected!")
+        return image, "Weed Severity: 0.00%", 0, 0
 
+    image = Image.fromarray(np.uint8(image))
     draw = ImageDraw.Draw(image)
+
     weed_count = 0
     crop_count = 0
 
@@ -44,7 +53,6 @@ def detect_weeds_and_classify(image):
             draw.text((x_min, y_min - 10), f"Crop: {confidence:.2f}", fill="green")
 
     severity = (weed_count / (weed_count + crop_count)) * 100 if (weed_count + crop_count) > 0 else 0
-
     return image, f"Weed Severity: {severity:.2f}%", weed_count, crop_count
 
 # Function to query Groq's Llama model for dynamic recommendations