在 Python 中使用 Flask 创建 ML 模型的 API

Question

我想使用烧瓶创建我的 ML 模型的 API。 该模型只是简单地获取图像并对其进行分类。 我想上传我的模型的 pickle 文件，并希望用户在运行时动态上传图像文件作为我的功能运行的参数。

以下是我尝试过的代码 sn-p，但存在一些问题，因为我无法从硬盘中检索文件：

from flask import Flask, request
import requests
import os
from tensorflow import keras as tfk
import matplotlib.pyplot as plt
from PIL import Image
from numpy import asarray
import pickle

app = Flask(__name__)


@app.route("/")
def hello():
    return "Image Classifier!"

@app.route('/', defaults={'path': ''})
@app.route('/image_reader/<path:image_path>/<pickle_file>', methods=['GET', 'POST'])
def image_classification(image_path):
    if request.json:
        data = request.get_json(image_path)

        image_path = data.get("image_path")
        print(image_path)
        r = requests.get(image_path, timeout=60)

        # save the image to disk
        temp_file = 'tmp/temp.jpg'
        f = open(temp_file, "wb")
        f.write(r.content)
        f.close()

        model = pickle.load(pickle_file)

        return image_path
    else:
        return "no json received"


if __name__ == '__main__':
    app.run(debug=True)

最初，我只是尝试同时采用上述两个参数，即上传 pickle_file 并将用户的动态文件路径作为参数。

Answer 1

当您上传文件时，它会存储在request.files 中，这是一个字典。

相反，你应该做的是这样的事情：

from flask import Flask, request
import requests
import os
from tensorflow import keras as tfk
import matplotlib.pyplot as plt
from PIL import Image
from numpy import asarray
import pickle

app = Flask(__name__)


@app.route("/")
def hello():
    return "Image Classifier!"

@app.route('/', defaults={'path': ''})
@app.route('/image_reader', methods=['POST'])
def image_classification():
    if request.method == 'POST':
        image_file = request.file.get('image_path')        
        pickle_file = request.file.get('pickle_file')

        # save the image to disk
        temp_file = 'tmp/temp.jpg'
        image_file.save(temp_file)

        # Save the pickle file to disk
        temp_picke_file = 'tmp/model.pickle'
        pickle_file.save(temp_pickle_file)

        # Load the model using the pickle file
        model = pickle.load(open(temp_pickle_file,'w'))

        return "Image saved to " + temp_file


if __name__ == '__main__':
    app.run(debug=True)