Update utils.py

utils.py

@@ -374,13 +374,21 @@ def predict(model, image, caption, box_threshold, text_threshold):
     return boxes, logits, phrases
 
 
-# utils.py (correct implementation)
-def predict_yolo(model, image_path, box_threshold):
-    result = model.predict(source=image_path, conf=box_threshold)  # ✅ Valid for YOLO
-    boxes = result[0].boxes.xyxy
-    conf = result[0].boxes.conf
-    phrases = [str(i) for i in range(len(boxes))]
-    return boxes, conf, phrases
+def predict_yolo(model, image_path, box_threshold, imgsz, scale_img, iou_threshold=0.7):
+    """Use YOLO model for object detection with correct parameters"""
+    kwargs = {
+        'conf': box_threshold,  # Correct confidence parameter
+        'iou': iou_threshold,   # Correct IoU parameter
+        'verbose': False
+    }
+    
+    if scale_img:
+        kwargs['imgsz'] = imgsz
+        
+    results = model.predict(image_path, **kwargs)
+    boxes = results[0].boxes.xyxy
+    conf = results[0].boxes.conf
+    return boxes, conf, [str(i) for i in range(len(boxes))]
 
 
 def get_som_labeled_img(img_path, model=None, BOX_TRESHOLD = 0.01, output_coord_in_ratio=False, ocr_bbox=None, text_scale=0.4, text_padding=5, draw_bbox_config=None, caption_model_processor=None, ocr_text=[], use_local_semantics=True, iou_threshold=0.9,prompt=None, scale_img=False, imgsz=None, batch_size=None):