Update app.py

app.py

@@ -12,30 +12,38 @@ model = VisionEncoderDecoderModel.from_pretrained("microsoft/trocr-base-stage1")
 
 # Enhance image before OCR
 def enhance_image(image):
-    image = image.convert("L")  # grayscale
-    image = ImageOps.invert(image)  # invert colors
-    image = ImageEnhance.Contrast(image).enhance(2)  # boost contrast
-    image = ImageEnhance.Sharpness(image).enhance(2)  # sharpen
-    image = image.resize((image.width * 2, image.height * 2))  # enlarge
+    # Convert to grayscale
+    image = image.convert("L")
+    # Invert (light text on dark bg works better)
+    image = ImageOps.invert(image)
+    # Increase contrast and sharpness
+    image = ImageEnhance.Contrast(image).enhance(2.0)
+    image = ImageEnhance.Sharpness(image).enhance(2.0)
+    # Resize (bigger = easier for OCR)
+    image = image.resize((image.width * 2, image.height * 2))
+    # Convert back to RGB for model compatibility
+    image = image.convert("RGB")
     return image
 
 # Extract weight
 def detect_weight(image):
     try:
         processed_image = enhance_image(image)
+
+        # Send to Hugging Face OCR model
         pixel_values = processor(images=processed_image, return_tensors="pt").pixel_values
         generated_ids = model.generate(pixel_values)
         generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
 
-        # Extract number
+        # Extract number using regex
         match = re.search(r"(\d{1,4}(?:\.\d{1,2})?)", generated_text)
         weight = match.group(1) if match else "Not detected"
 
-        # Timestamp
+        # Get current IST time
         ist = pytz.timezone('Asia/Kolkata')
         current_time = datetime.now(ist).strftime("%Y-%m-%d %H:%M:%S")
 
-        return f"Weight: {weight} kg\nCaptured At: {current_time}", image
+        return f"Weight: {weight} kg\nCaptured At: {current_time} (IST)", image
     except Exception as e:
         return f"Error: {str(e)}", image
 
@@ -45,7 +53,7 @@ interface = gr.Interface(
     inputs=gr.Image(type="pil", label="Upload or Capture Image"),
     outputs=[gr.Textbox(label="Weight Info"), gr.Image(label="Snapshot")],
     title="⚖️ Auto Weight Detector (No Tesseract)",
-    description="Detects weight from digital scale image using Hugging Face TrOCR."
+    description="Detects weight from digital scale image using Hugging Face TrOCR. Shows weight and capture time (IST)."
 )
 
 interface.launch()