谷歌云服务器在查询对象本地化时被迫关闭

Question

我有时会收到以下错误信息

google.api_core.exceptions.ServiceUnavailable: 503 Getting metadata from plugin failed with error: ('Connection aborted.', ConnectionResetError(10054, 'An existing connection was forcibly closed by the remote host', None, 10054, None))

运行后

# server/classifiers/gcloud/identifier.py


import io
import os
import json
from google.cloud import vision
from google.cloud.vision import types
from collections import Counter

client = vision.ImageAnnotatorClient()

with io.open("config/labels.json", "r") as f:
    LABELS = json.loads(f.read())


def identify_from_string(blob):
    image = types.Image(content=blob)
    response = client.object_localization(image=image)
    labels = response.localized_object_annotations
    objects = set(label.name for label in labels)

    c = Counter()
    for label, s in LABELS.items():
        for ob in objects:
            if ob in s:
                c[label] += s[ob]

    if not c: print("Sorry, the server is currently full.")
    return str(c)

抛出的错误与 response = client.object_localization(image=image). 在我的 app.py我有

# server/app.py

import os
import json
from server.classifiers.gcloud import identifier
from flask import Flask, Response, render_template, send_file, request

@app.route('/classify', methods=['POST'])
def classify():
    app.logger.info("Got image to /classify")
    file = request.files['image']
    blob = file.read()
    results = identifier.identify_from_string(blob)
    return Response(response=json.dumps(dict(response)), status=200)

其中server.classifiers.gcloud.identifier指向上面的文件。偶尔，这个问题会突然消失。有什么办法解决吗？

Answer 1

看来你很有可能需要下载服务账户凭证，并导出环境变量以及你的GOOGLE_CLOUD_PROJECT，例如在*nix上。

export GOOGLE_CLOUD_PROJECT=blue-jet-123
export GOOGLE_APPLICATION_CREDENTIALS=$HOME/path/to/service.json

我想重现这个错误，但我的flask服务器没有启动，你能不能说明configlabels.json的内容，你可以把你的文件夹树化，以确保我们有相同的目录结构。