Gradio：Pythonでインタラクティブな機械学習アプリケーションを簡単に作成するためのフレームワーク

pip install gradio

import gradio as gr

def echo(text):
    return text

iface = gr.Interface(fn=echo, inputs="text", outputs="text", title="Echo App", description="Enter some text and see the output.")

iface.launch()

python app.py

pip install tensorflow

import ssl
ssl._create_default_https_context = ssl._create_unverified_context

import gradio as gr
import tensorflow as tf
import numpy as np

# MobileNetモデルのロード
model = tf.keras.applications.MobileNetV2(weights="imagenet")

def classify_image(image):
    # 画像の前処理
    image = tf.image.resize(image, (224, 224))
    image = tf.keras.applications.mobilenet_v2.preprocess_input(image)
    image = np.expand_dims(image, axis=0)
    
    # 画像分類
    preds = model.predict(image)
    decoded_preds = tf.keras.applications.mobilenet_v2.decode_predictions(preds, top=5)[0]
    return {label: float(score) for (_, label, score) in decoded_preds}

iface = gr.Interface(fn=classify_image, inputs="image", outputs="label", title="Image Classification", description="Upload an image to classify.")

iface.launch()

pip install transformers
pip install torch

import gradio as gr
from transformers import Wav2Vec2ForCTC, Wav2Vec2Tokenizer
import torch
import numpy as np

# モデルとトークナイザーのロード
tokenizer = Wav2Vec2Tokenizer.from_pretrained("facebook/wav2vec2-base-960h")
model = Wav2Vec2ForCTC.from_pretrained("facebook/wav2vec2-base-960h")

def transcribe_audio(audio):
    # 音声データの前処理
    audio = np.mean(audio, axis=1)  # 音声データがステレオの場合はモノラルに変換
    input_values = tokenizer(audio, return_tensors="pt", sampling_rate=16000).input_values
    
    # 音声認識
    with torch.no_grad():
        logits = model(input_values).logits
    predicted_ids = torch.argmax(logits, dim=-1)
    transcription = tokenizer.batch_decode(predicted_ids)[0]
    return transcription

iface = gr.Interface(fn=transcribe_audio, inputs="microphone", outputs="text", title="Speech Recognition", description="Speak into the microphone to transcribe speech to text.")

iface.launch()