app.py

from transformers import Qwen2_5_VLForConditionalGeneration, AutoProcessor
import gradio as gr

# 加载模型和处理器
model = Qwen2_5_VLForConditionalGeneration.from_pretrained("Qwen/Qwen2.5-VL-7B-Instruct", device_map="auto")
processor = AutoProcessor.from_pretrained("Qwen/Qwen2.5-VL-7B-Instruct")

# 定义识别和分析函数
def recognize_and_analyze(image, text_prompt):
    messages = [
        {
            "role": "user",
            "content": [
                {"type": "image", "image": image},
                {"type": "text", "text": text_prompt},
            ],
        }
    ]

    # 处理视觉信息
    inputs = processor(images=image, return_tensors="pt")
    inputs = inputs.to(model.device)

    # 生成输出
    generated_ids = model.generate(**inputs)
    output_text = processor.batch_decode(generated_ids, skip_special_tokens=True)
    return output_text[0]

# 设置Gradio界面
interface = gr.Interface(
    fn=recognize_and_analyze,
    inputs=[
        gr.inputs.Image(type="filepath", label="上传图像"),
        gr.inputs.Textbox(label="输入描述文本"),
    ],
    outputs=gr.outputs.Textbox(label="识别结果"),
    title="Qwen2.5-VL 物体识别与分析",
    description="上传图像并输入描述文本以获取识别和分析结果。",
)

# 启动Gradio应用
if __name__ == "__main__":
    interface.launch()