.

app.py

@@ -9,7 +9,7 @@ from utils import Image2Text
 @spaces.GPU(duration=15)
 def greet(input_img):
     global image_to_text
-    contents = image_to_text.get_text([input_img], num_beams=4)
+    contents = image_to_text.get_text(input_img, num_beams=4)
     return '\n'.join(contents)
 
 examples_path = os.path.dirname(__file__)

utils.py

@@ -51,11 +51,8 @@ class Image2Text:
         model.eval()
         return model, processor
 
-    def load_img(self, img_path, width=480, height=480):
-        if type(img_path) == np.ndarray:
-            image = Image.fromarray(img_path)
-        else:
-            image = Image.open(img_path)
+    def load_img(self, inputs, width=480, height=480):
+        image = Image.fromarray(inputs)
         image = aspect_ratio_preserving_resize_and_crop(image, target_width=width, target_height=height).convert("RGB")
         img = self.processor(image , return_tensors="pt", size=(width, height)).pixel_values
         pixel_values = img.to(self.device)