如何使用请求模块和 Flask 通过 http post 请求发送 numpy 数组或 pytorch 张量

Question

我有一张图片，我想将它发送到服务器。我正在使用 requests 模块来执行简单的 post 请求，如下所示（信息是字典）：

    import requests

    print(type(info["array_image"]))
    print(type(info["visual_features"]))
    response = requests.post("url", data=info)

输出：

<class 'numpy.ndarray'>
<class 'torch.Tensor'>

在服务器端，我试图至少将它们作为数组接收：

from flask import Flask, request

@app.route('/path', methods=['POST'])
def function_name():
    visual_features = request.form['visual_features']
    array_image = request.form['array_image']
    print(type(array_image))
    print(type(visual_features))

输出：

<class 'str'>
<class 'str'>

我想获取一个字节数组来构建图像，但我得到的是一个字符串... 如果我找不到方法，我将在 bas64 中对数组进行编码，然后在服务器中对其进行解码...

Answer 1

您可以尝试将数据转换为正确的 JSON 或使用 Python pickle module or if you have an image as you've mentioned you can send it as a file (multipart request) to the server as shown in the examples here.

Answer 2

感谢@praba230890 给我一个简单易懂的例子。

我仍然会在这里写下解决方案，因为提供的 link 不完全适合我的情况。

import pickle
import io

bytes_image = pickle.dumps(info["array_image"])
stream = io.BytesIO(bytes_image)
files = {"bytes_image": stream}

info["array_image"] = None

response = http.post("url", data=info, files=files)

在服务器端：

from flask import Flask, request

@app.route('/path', methods=['POST'])
def function_name():
    image = request.files.get('bytes_image')
    bytes_image = image.read()

如果您想从文件中获取图像，则：

requests.post("http://localhost:5000/predict",
                 files={"file": open('<PATH/TO/.jpg/FILE>/cat.jpg','rb')})

我目前使用的解决方案：

记住info["array_image"]是一个numpy数组，info是一个字典

    import io
    info["image_shape_width"] = info["array_image"].shape[0]
    info["image_shape_height"] = info["array_image"].shape[1]

    bytes_image = info["array_image"].tobytes()
    stream = io.BytesIO(bytes_image)
    files = {"bytes_image": stream}

    info["array_image"] = None

    response = http.post(self.ip + "path", data=info, files=files)

然后收到

    from flask import Flask, request
    import numpy as np

    @app.route('/path', methods=['POST'])
    def function_name():    
        bytes_image = request.files.get('bytes_image')
        bytes_image = bytes_image.read()
        array_image = np.frombuffer(bytes_image, dtype=dtype)
        shape = (int(request.form['image_shape_width']), int(request.form['image_shape_height']), 3)         
        array_image = np.reshape(array_image, shape)
                                       
        image = Image.fromarray(array_image)

如何使用请求模块和 Flask 通过 http post 请求发送 numpy 数组或 pytorch 张量

how to send an numpy array or a pytorch Tensor through http post request using requests module and Flask

python

numpy

flask

python-requests

pytorch

我目前使用的解决方案：