app.py

import gradio as gr
from transformers import pipeline

# Hugging Faceの互換性のあるモデルをロード（image-to-textタスク用）
model_name = "Salesforce/blip-image-captioning-base"
image_to_text = pipeline("image-to-text", model=model_name)

# Gradioの関数定義
def generate_text_from_image(image):
    # 画像からテキストを生成
    result = image_to_text(image)
    return result[0]["generated_text"]

# Gradioインターフェースの設定
iface = gr.Interface(
    fn=generate_text_from_image,
    inputs=gr.Image(type="pil"),
    outputs="text",
    title="Image to Text with BLIP Model",
    description="Upload an image to get a descriptive text generated by the BLIP image captioning model."
)

# Gradioアプリケーションの起動
iface.launch(server_name="0.0.0.0", server_port=7860)