ollama接入图像识别大模型

安装ollama和qwen2.5-vl

安装ollama

ollama run qwen2.5vl

控制台调用

api调用

api

请求地址http://localhost:11434/api/chat
请求内容

{"model": "qwen2.5vl","stream": false,"messages": [{"role": "user","content": "有效期是多少","images": ["图片的base64编码"]}]
}

python

cd /home/wjl/python-project
mkdir ollamastu
cd ollamastu
uv venv -p3.12 .venv
uv inituv add requests

把图片复制过去 cp /mnt/e/Desktop/001.jpg /home/wjl/python-project/ollamastu/，然后修改mian.py代码

import requests
import base64
import jsonif __name__=="__main__":# 图片转Base64with open("001.jpg", "rb") as image_file:image_data = base64.b64encode(image_file.read()).decode('utf-8')# API请求参数payload = {"model": "qwen2.5vl","prompt": "注册证编号是多少","images": [image_data]}# 发送请求response = requests.post("http://localhost:11434/api/generate",json=payload,stream=True)# 处理响应for chunk in response.iter_lines():if chunk:print(json.loads(chunk)["response"])

参考

https://blog.csdn.net/csdn122345/article/details/145701484
https://blog.csdn.net/weixin_42357472/article/details/146221932