diff --git a/__pycache__/utils.cpython-312.pyc b/__pycache__/utils.cpython-312.pyc
index 79598e5..a1cfe48 100644
Binary files a/__pycache__/utils.cpython-312.pyc and b/__pycache__/utils.cpython-312.pyc differ
diff --git a/gradio_demo.py b/gradio_demo.py
index 14dd6d0..8711c57 100644
--- a/gradio_demo.py
+++ b/gradio_demo.py
@@ -12,7 +12,8 @@ from utils import check_ocr_box, get_yolo_model, get_caption_model_processor, ge
 import torch
 from PIL import Image
 
-yolo_model = get_yolo_model(model_path='weights/icon_detect/best.pt')
+# yolo_model = get_yolo_model(model_path='weights/icon_detect/best.pt')
+yolo_model = get_yolo_model(model_path='weights/icon_detect_v1_5/best.pt')
 caption_model_processor = get_caption_model_processor(model_name="florence2", model_name_or_path="weights/icon_caption_florence")
 # caption_model_processor = get_caption_model_processor(model_name="blip2", model_name_or_path="weights/icon_caption_blip2")
 
@@ -57,10 +58,11 @@ def process(
     ocr_bbox_rslt, is_goal_filtered = check_ocr_box(image_save_path, display_img = False, output_bb_format='xyxy', goal_filtering=None, easyocr_args={'paragraph': False, 'text_threshold':0.9}, use_paddleocr=use_paddleocr)
     text, ocr_bbox = ocr_bbox_rslt
     # print('prompt:', prompt)
-    dino_labled_img, label_coordinates, parsed_content_list = get_som_labeled_img(image_save_path, yolo_model, BOX_TRESHOLD = box_threshold, output_coord_in_ratio=True, ocr_bbox=ocr_bbox,draw_bbox_config=draw_bbox_config, caption_model_processor=caption_model_processor, ocr_text=text,iou_threshold=iou_threshold, imgsz=imgsz)  
+    dino_labled_img, label_coordinates, parsed_content_list = get_som_labeled_img(image_save_path, yolo_model, BOX_TRESHOLD = box_threshold, output_coord_in_ratio=True, ocr_bbox=ocr_bbox,draw_bbox_config=draw_bbox_config, caption_model_processor=caption_model_processor, ocr_text=text,iou_threshold=iou_threshold, imgsz=imgsz,)  
     image = Image.open(io.BytesIO(base64.b64decode(dino_labled_img)))
     print('finish processing')
-    parsed_content_list = '\n'.join(parsed_content_list)
+    parsed_content_list = '\n'.join([f'icon {i}: ' + str(v) for i,v in enumerate(parsed_content_list)])
+    # parsed_content_list = str(parsed_content_list)
     return image, str(parsed_content_list)
 
 
diff --git a/imgs/demo_image.jpg b/imgs/demo_image.jpg
index 4620e66..ab00b9e 100644
Binary files a/imgs/demo_image.jpg and b/imgs/demo_image.jpg differ
diff --git a/imgs/demo_image_som.jpg b/imgs/demo_image_som.jpg
index 538855c..0156c81 100644
Binary files a/imgs/demo_image_som.jpg and b/imgs/demo_image_som.jpg differ
diff --git a/imgs/saved_image_demo.png b/imgs/saved_image_demo.png
index 1719c03..feaff69 100644
Binary files a/imgs/saved_image_demo.png and b/imgs/saved_image_demo.png differ
diff --git a/utils.py b/utils.py
index c88ce6a..c4865b0 100755
--- a/utils.py
+++ b/utils.py
@@ -75,7 +75,7 @@ def get_yolo_model(model_path):
 
 
 @torch.inference_mode()
-def get_parsed_content_icon(filtered_boxes, starting_idx, image_source, caption_model_processor, prompt=None, batch_size=32):
+def get_parsed_content_icon(filtered_boxes, starting_idx, image_source, caption_model_processor, prompt=None, batch_size=None):
     # Number of samples per batch, --> 256 roughly takes 23 GB of GPU memory for florence model
 
     to_pil = ToPILImage()
@@ -99,6 +99,7 @@ def get_parsed_content_icon(filtered_boxes, starting_idx, image_source, caption_
     
     generated_texts = []
     device = model.device
+    # batch_size = 64
     for i in range(0, len(croped_pil_image), batch_size):
         start = time.time()
         batch = croped_pil_image[i:i+batch_size]
@@ -398,7 +399,7 @@ def predict_yolo(model, image_path, box_threshold, imgsz, scale_img, iou_thresho
     return boxes, conf, phrases
 
 
-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):
+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=64):
     """ ocr_bbox: list of xyxy format bbox
     """
     image_source = Image.open(img_path).convert("RGB")
@@ -432,6 +433,7 @@ def get_som_labeled_img(img_path, model=None, BOX_TRESHOLD = 0.01, output_coord_
     # get the index of the first 'content': None
     starting_idx = next((i for i, box in enumerate(filtered_boxes_elem) if box['content'] is None), -1)
     filtered_boxes = torch.tensor([box['bbox'] for box in filtered_boxes_elem])
+    print('len(filtered_boxes):', len(filtered_boxes), starting_idx)
 
     
     # get parsed icon local semantics
@@ -501,7 +503,7 @@ def check_ocr_box(image_path, display_img = True, output_bb_format='xywh', goal_
         else:
             text_threshold = easyocr_args['text_threshold']
         result = paddle_ocr.ocr(image_path, cls=False)[0]
-        conf = [item[1] for item in result]
+        # conf = [item[1] for item in result]
         coord = [item[0] for item in result if item[1][1] > text_threshold]
         text = [item[1][0] for item in result if item[1][1] > text_threshold]
     else:  # EasyOCR