Create app.py

app.py

@@ -0,0 +1,67 @@
+import gradio as gr
+import tensorflow as tf
+import numpy as np
+import cv2
+from PIL import Image, ImageOps
+
+def load_model():
+    try:
+        return tf.keras.models.load_model("mnist_model.h5")
+    except:
+        model = tf.keras.models.Sequential([
+            tf.keras.layers.Flatten(input_shape=(28, 28)),
+            tf.keras.layers.Dense(128, activation='relu'),
+            tf.keras.layers.Dense(10, activation='softmax')
+        ])
+        model.compile(optimizer='adam',
+                      loss='sparse_categorical_crossentropy',
+                      metrics=['accuracy'])
+        (x_train, y_train), _ = tf.keras.datasets.mnist.load_data()
+        x_train = x_train / 255.0
+        model.fit(x_train, y_train, epochs=5, batch_size=32, verbose=0)
+        model.save("mnist_model.h5")
+        return model
+
+model = load_model()
+
+def segment_digits(image: Image.Image):
+    img = np.array(image.convert("L"))
+    img = ImageOps.invert(Image.fromarray(img))
+    img = np.array(img)
+    img = cv2.resize(img, (img.shape[1]*2, img.shape[0]*2))
+    _, thresh = cv2.threshold(img, 128, 255, cv2.THRESH_BINARY)
+    contours, _ = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
+
+    digit_images = []
+    for cnt in sorted(contours, key=lambda x: cv2.boundingRect(x)[0]):
+        x, y, w, h = cv2.boundingRect(cnt)
+        digit = thresh[y:y+h, x:x+w]
+        digit = cv2.resize(digit, (28, 28))
+        digit = digit / 255.0
+        digit_images.append(digit.reshape(1, 28, 28))
+    return digit_images
+
+def classify_multi_digit(image):
+    digits = segment_digits(image)
+    result = ""
+    confidences = {}
+    for i, digit in enumerate(digits):
+        pred = model.predict(digit)[0]
+        digit_class = np.argmax(pred)
+        result += str(digit_class)
+        confidences[f"Digit {i+1} ({digit_class})"] = round(np.max(pred), 2)
+    return f"Predicted Number: {result}", confidences
+
+demo = gr.Interface(
+    fn=classify_multi_digit,
+    inputs=gr.Image(type="pil", label="Upload image with digits (e.g. 178)"),
+    outputs=[
+        gr.Text(label="Predicted Number"),
+        gr.Label(label="Confidence per Digit")
+    ],
+    title="🧠 Multi-digit MNIST Classifier",
+    description="Upload an image containing multiple handwritten digits (e.g. '178'). The app segments and classifies each digit using a simple MNIST-trained neural network."
+)
+
+if __name__ == "__main__":
+    demo.launch()