如何将通过 post 请求发送的图像发送到模型进行预测

Question

我是运行一个 Flask 应用程序，它通过 post 请求接收 base64 编码的图像，并应将其发送以进行分类。不幸的是，这是我遇到错误的地方。这是我发出 post 请求的代码：

def encode_image(img_path):

    with open(img_path, 'rb') as f:

        encoded_string = base64.b64encode(f.read())

    return encoded_string


def send_image(img_path):

    img = encode_image(img_path)

    data = {'imageBase64': img.decode('UTF-8')}

    r = requests.post(url, data = json.dumps(data))

send_image('/path/to/image.jpg')

这是我的烧瓶应用程序代码：

@app.route('/', methods = ['POST'])
def predict():
    if request.method == 'POST':  
        base64img = request.data
    
        base64img += b'=' * (-len(base64img) % 4)
    
        img = base64.b64decode(base64img)

        img = image.load_img(BytesIO(img), target_size = (28, 28), color_mode = 'grayscale')
        img = Image.open(img)

        img = image.img_to_array(image.load_img(BytesIO(img)), target_size = (28, 28))
        payload = {
            'instances': [{'input_image': img.to_list()}]
        }

        r = request.post(url, json = payload)
        pred = json.loads(r.content.decode('utf-8'))

        return pred

第 img = image.load_img(BytesIO(img)) 行抛出此错误：

TypeError: expected str, bytes or os.PathLike object, not BytesIO

我不确定如何解决这个问题，有人有什么想法吗？

Answer 1

好的，我设法做到了：

@app.route('/', methods = ['POST'])
def predict():
    if request.method == 'POST':
        base64img = request.get_json()['img'].encode('utf-8')

        img = base64.b64decode(base64img)
        with open('imgtemp.png', 'wb') as f:
            f.write(img)

        img = cv2.imread('imgtemp.png', 1)
    
        img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)

        payload = {
            'instances': [{'input_image': img.tolist()}]
        }

        r = requests.post(url, json = payload)
        pred = json.loads(r.content.decode('utf-8'))

        return pred

如何将通过 post 请求发送的图像发送到模型进行预测

How to send an image that was sent with a post request to a model for prediction

python

tensorflow

flask

request

tensorflow-serving