from streamlit_drawable_canvas import st_canvas import tensorflow as tf import streamlit as st import numpy as np import cv2 # Page Configuration st.set_page_config(page_title="Digit Recognition", page_icon="🔢", layout="wide") # Modern styling st.markdown(""" """, unsafe_allow_html=True) # Header st.title("🔢 Handwritten Digit Recognition") st.markdown("Draw a **digit (0-9)** on the black canvas, and let the model predict it.") # Cache Model @st.cache_resource def load_model(): try: return tf.keras.models.load_model('model.keras') except Exception as e: st.error(f"Error loading model: {e}") return None model = load_model() # Layout col1, col2 = st.columns([1, 1], gap="large") with col1: st.subheader("Drawing Area") canvas_result = st_canvas( fill_color="rgba(255, 165, 0, 0.3)", stroke_width=15, stroke_color="#FFFFFF", background_color="#000000", height=280, width=280, drawing_mode="freedraw", display_toolbar=True, key="canvas", ) with col2: st.subheader("Prediction") if canvas_result.image_data is not None: img = canvas_result.image_data if np.sum(img) > 0: img = cv2.cvtColor(img, cv2.COLOR_RGBA2GRAY) img_resized = cv2.resize(img, (28, 28), interpolation=cv2.INTER_AREA) img_normalized = img_resized / 255.0 img_reshaped = img_normalized.reshape(1, 28, 28, 1) prediction = model.predict(img_reshaped) predicted_class = np.argmax(prediction) confidence = np.max(prediction) st.markdown(f"

{predicted_class}

", unsafe_allow_html=True) st.success(f"Confidence: {confidence * 100:.2f}%") st.write("Model Input (28x28):") st.image(img_resized, width=100) else: st.info("Please draw a digit.")