Update app.py

app.py

@@ -16,8 +16,8 @@ model = VisionEncoderDecoderModel.from_pretrained('microsoft/trocr-large-handwri
 def display_sketch(sketch):
     logging.debug(f"Received sketch data: {sketch}")
 
-    if isinstance(sketch, dict) and "composite" in sketch:
-        image_data = sketch["composite"]
+    if isinstance(sketch, np.ndarray):
+        image_data = sketch
         logging.debug(f"Image data type: {type(image_data)}")
         logging.debug(f"Image data shape: {np.array(image_data).shape}")
 
@@ -25,7 +25,7 @@ def display_sketch(sketch):
         plt.axis('off')
         
         with tempfile.NamedTemporaryFile(delete=False, suffix=".png") as temp_file:
-            plt.savefig(temp_file.name, bbox_inches='tight')
+            plt.savefig(temp_file.name, bbox_inches='tight', pad_inches=0)
             temp_file_path = temp_file.name
         
         return temp_file_path
@@ -35,14 +35,18 @@ def display_sketch(sketch):
         return error_message
 
 def recognize_text(image_path):
-    # Load the image
-    image = Image.open(image_path).convert("RGB")
-    # Prepare the image for the model
-    pixel_values = processor(image, return_tensors="pt").pixel_values
-    # Generate the text
-    generated_ids = model.generate(pixel_values)
-    generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
-    return generated_text
+    try:
+        # Load the image
+        image = Image.open(image_path).convert("RGB")
+        # Prepare the image for the model
+        pixel_values = processor(image, return_tensors="pt").pixel_values
+        # Generate the text
+        generated_ids = model.generate(pixel_values)
+        generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
+        return generated_text
+    except Exception as e:
+        logging.error(f"Error in recognizing text: {e}")
+        return "Error in recognizing text"
 
 with gr.Blocks() as demo:
     sketchpad = gr.Sketchpad(label="Draw Something")