Gradio와 Steamlit의 코드 및 인터페이스 비교

Gradio

import tensorflow as tf  
import numpy as np  
import gradio as gr  
  
# MNIST 데이터 로드  
(x_train, y_train), (x_test, y_test) = tf.keras.datasets.mnist.load_data()  
  
# 데이터 전처리  
x_train = x_train.reshape(-1, 28, 28, 1).astype('float32') / 255  
x_test = x_test.reshape(-1, 28, 28, 1).astype('float32') / 255  
  
# 모델 정의  
def create_model():  
    model = tf.keras.Sequential([  
        tf.keras.layers.Conv2D(32, (3, 3), activation='relu', input_shape=(28, 28, 1)),  
        tf.keras.layers.MaxPooling2D((2, 2)),  
        tf.keras.layers.Flatten(),  
        tf.keras.layers.Dense(128, activation='relu'),  
        tf.keras.layers.Dense(10, activation='softmax')  
    ])  
    model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])  
    return model  
  
model = create_model()  
  
# 학습 함수  
def train_model():  
    class AccuracyCallback(tf.keras.callbacks.Callback):  
        def __init__(self):  
            self.accuracies = []  
  
        def on_epoch_end(self, epoch, logs=None):  
            self.accuracies.append(logs['accuracy'])  
  
    accuracy_callback = AccuracyCallback()  
  
    history = model.fit(x_train, y_train, epochs=3, validation_split=0.1, verbose=0,   
                        callbacks=[accuracy_callback])  
      
    accuracy_text = "Training accuracies:\n"  
    for epoch, acc in enumerate(accuracy_callback.accuracies, 1):  
        accuracy_text += f"Epoch {epoch}: {acc:.4f}\n"  
    accuracy_text += f"\nFinal accuracy: {accuracy_callback.accuracies[-1]:.4f}"  
    return accuracy_text  
  
# 예측 함수  
def predict():  
    idx = np.random.randint(0, len(x_test))  
    image = x_test[idx]  
    true_label = y_test[idx]  
      
    prediction = model.predict(image[np.newaxis, ...])[0]  
    predicted_label = np.argmax(prediction)  
      
    display_image = (image.reshape(28, 28) * 255).astype(np.uint8)  
      
    return (f"Predicted: {predicted_label}, True: {true_label}", display_image)  
  
# Gradio 인터페이스  
with gr.Blocks() as demo:  
    gr.Markdown("# MNIST 분류기")  
      
    with gr.Row():  
        train_button = gr.Button("학습")  
        output_text = gr.Textbox(label="학습 결과", lines=7)  
      
    with gr.Row():  
        predict_button = gr.Button("예측")  
        output_label = gr.Textbox(label="예측 결과")  
        output_image = gr.Image(label="예측 이미지")  
      
    train_button.click(train_model, outputs=output_text)  
    predict_button.click(predict, outputs=[output_label, output_image])  
  
demo.launch()

Streamlit

import streamlit as st  
import tensorflow as tf  
import numpy as np  
import matplotlib.pyplot as plt  
  
# MNIST 데이터 로드  
@st.cache_resource  
def load_data():  
    return tf.keras.datasets.mnist.load_data()  
  
(x_train, y_train), (x_test, y_test) = load_data()  
  
# 데이터 전처리  
x_train = x_train.reshape(-1, 28, 28, 1).astype('float32') / 255  
x_test = x_test.reshape(-1, 28, 28, 1).astype('float32') / 255  
  
# 모델 정의  
@st.cache_resource  
def create_model():  
    model = tf.keras.Sequential([  
        tf.keras.layers.Conv2D(32, (3, 3), activation='relu', input_shape=(28, 28, 1)),  
        tf.keras.layers.MaxPooling2D((2, 2)),  
        tf.keras.layers.Flatten(),  
        tf.keras.layers.Dense(128, activation='relu'),  
        tf.keras.layers.Dense(10, activation='softmax')  
    ])  
    model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])  
    return model  
  
model = create_model()  
  
# 학습 함수  
def train_model():  
    class AccuracyCallback(tf.keras.callbacks.Callback):  
        def on_epoch_end(self, epoch, logs=None):  
            accuracy = logs.get('accuracy')  
            st.session_state.accuracies.append(accuracy)  
            st.session_state.progress_bar.progress((epoch + 1) / 5)  
            st.session_state.status_text.text(f"Epoch {epoch + 1}/5 completed")  
  
    accuracy_callback = AccuracyCallback()  
      
    history = model.fit(x_train, y_train, epochs=5, validation_split=0.1, verbose=0,   
                        callbacks=[accuracy_callback])  
      
    st.session_state.trained = True  
  
# 예측 함수  
def predict():  
    idx = np.random.randint(0, len(x_test))  
    image = x_test[idx]  
    true_label = y_test[idx]  
      
    prediction = model.predict(image[np.newaxis, ...])[0]  
    predicted_label = np.argmax(prediction)  
      
    return image, predicted_label, true_label  
  
# Streamlit 앱  
st.title("MNIST 분류기")  
  
if 'trained' not in st.session_state:  
    st.session_state.trained = False  
    st.session_state.accuracies = []  
  
col1, col2 = st.columns(2)  
  
with col1:  
    if st.button("학습"):  
        st.session_state.accuracies = []  
        st.session_state.progress_bar = st.progress(0)  
        st.session_state.status_text = st.empty()  
        train_model()  
        st.session_state.progress_bar.empty()  
        st.session_state.status_text.empty()  
          
        accuracy_text = "Training accuracies:\n"  
        for epoch, acc in enumerate(st.session_state.accuracies, 1):  
            accuracy_text += f"Epoch {epoch}: {acc:.4f}\n"  
        accuracy_text += f"\nFinal accuracy: {st.session_state.accuracies[-1]:.4f}"  
        st.text_area("학습 결과", accuracy_text, height=200)  
  
with col2:  
    if st.button("예측"):  
        if not st.session_state.trained:  
            st.warning("먼저 모델을 학습시켜주세요.")  
        else:  
            image, predicted_label, true_label = predict()  
            st.image(image.reshape(28, 28), caption=f"Predicted: {predicted_label}, True: {true_label}", width=200)  
  
st.markdown("---")  
st.markdown("학습 버튼을 눌러 모델을 학습시킨 후, 예측 버튼을 눌러 랜덤한 MNIST 이미지에 대한 예측을 확인하세요.")

• 학습시에는 Progess Bar로 진척도를 보여주다가 종료되면 Epoch별 Metric을 보여주었는데, 예측버튼 클릭시이 부분이 사라졌다