[js]#!/usr/bin/env python
# -*- coding: utf-8 -*-
# ------------------------------------------------------------------------------
# フローデザイナーのモデルジェネレーター予測（オンライン）ブロックで画像分類タイプをしている
# フローをAPIから起動して終了確認して結果の変数値を取得する
#
#   フローデザイナーから下記3点を確認し設定する
#     フローデザイナーのサービス設定で取得したAPIトークン
#     フローデザイナーのURLの固有部分
#     実行するフローに設定したID
#     JPEGファイルパス
#
#  　実行サンプル
#     python ./img_prd.py <blocks_designer> <blocks_token> <blocks_flow> '/dir/img.jpg'
#       ※JSON文字列は "_" を最初に定義しておくと、そのままフローで "_"で利用できる。
#
# ------------------------------------------------------------------------------
import sys
import time
import base64
import urllib.request, json
from distutils.util import strtobool

if __name__ == '__main__':

    # 引数取得
    args = sys.argv

    # 引数チェック
    if len(args) != 5:
        # Usage出力
        print("Please call with collect arguments.")
        print("  python ./img_prd.py <blocks_designer> <blocks_token> <blocks_flow> '/dir/img.jpg'")
        quit()

    # BLOCKS実行フロー情報
    blocks_designer = args[1]
    blocks_token = args[2]
    blocks_flow = args[3]

    # 画像ファイルパス
    image_file = args[4]

    print("--------------------------------------------------------------------------------")
    print("Flow Designer flow Starting")
    print("Designer URL : " + blocks_designer)
    print("Flow ID : " + blocks_flow)
    print("JPEG FILE : " + image_file)

    # --------------------------------------------------------------------------
    # 画像ファイルをBASE64化してJSON作成
    # --------------------------------------------------------------------------
    b64 = base64.encodestring(open(image_file, 'rb').read())
    blocks_json = {
        "_" :
            {
                "key" : image_file,
                "image" : {
                    "b64" : b64.decode('utf8')
                }
            }
        }
    json_data = json.dumps(blocks_json).encode("utf-8")

    # --------------------------------------------------------------------------
    # BLOCKS フローデザイナーフロー起動
    # --------------------------------------------------------------------------
    url = "https://" + blocks_designer + ".magellanic-clouds.net/flows/" + blocks_flow + ".json"
    method = "POST"
    headers = {
        "Content-Type" : "application/json" ,
        "Authorization" : "Bearer " + blocks_token
    }

    # httpリクエストを準備してPOST
    request = urllib.request.Request(url, data=json_data, method=method, headers=headers)
    with urllib.request.urlopen(request) as response:
        response_body = response.read().decode("utf-8")
        json_obj = json.loads(response_body)
        result = json_obj['result']
        job_id = json_obj['job_id']

        if result == True:
            print("flow starting is successed.")
        else:
            print("flow starting is failed.")
            exit()

    # --------------------------------------------------------------------------
    # BLOCKS フローデザイナー実行ステータス取得
    # --------------------------------------------------------------------------
    status = ""
    while status not in ["finished" ,"failed", "canceled"] :

        # 1秒待つ
        time.sleep(1)

        # リクエスト情報作成
        url = "https://" + blocks_designer + ".magellanic-clouds.net/flows/" + blocks_flow + "/jobs/" + str(job_id) + "/status.json"
        method = "GET"
        # httpリクエストを準備してGET
        request = urllib.request.Request(url,  method=method, headers=headers)
        with urllib.request.urlopen(request) as response:
            response_body = response.read().decode("utf-8")

            json_obj = json.loads(response_body)
            status = json_obj['status']
            print("job_id " + str(job_id) + " is " + status)

    if status not in ["failed","canceled"]:
        print("flow is successed.")
    else:
        print("flow is failed.")
        exit()

    # --------------------------------------------------------------------------
    # BLOCKS フローデザイナー変数値取得
    # --------------------------------------------------------------------------
    # リクエスト情報作成
    url = "https://" + blocks_designer + ".magellanic-clouds.net/flows/" + blocks_flow + "/jobs/" + str(job_id) + "/variable.json"
    method = "GET"
    # httpリクエストを準備してGET
    request = urllib.request.Request(url,  method=method, headers=headers)
    with urllib.request.urlopen(request) as response:
        response_body = response.read().decode("utf-8")

        json_obj = json.loads(response_body)
        print(json_obj)
[/js]

[js]
# -*- coding: utf-8 -*-
# ------------------------------------------------------------------------------
# 画像分類の予測をモデルジェネレーターで作成したモデルでローカル環境で予測する
#
#  　前提：下記のフォルダ構成に配置する
#     /base_dir
#       /predict_images.py（本プログラム）
#       /blocks_ml_99999
#         /saved_model.pb（ダウンロードしたモデル）
#
#  　実行イメージ
#     python ./predict_image.py <images_dir>
#
# ------------------------------------------------------------------------------
import os
import sys
import time
import tensorflow as tf
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
saved_model_dir = "blocks_ml_99999"

if __name__ == '__main__':

    # 引数取得
    args = sys.argv
    if len(args) != 2:
        print("argments error!")
        print("python ./predict_images.py <images_dir>")
        quit()

    images_dir = args[1]
    ext_list = ["jpg", "jpeg","JPG","JPEG"]

    print("="*80)
    before_get_files = time.time()

    file_list = []
    for root, dirs, files in os.walk(images_dir):
        for ext in ext_list:
            file_list.extend([os.path.join(root, file) for file in files if ext in file])

    after_get_files = time.time()
    elapsed_get_files = after_get_files - before_get_files
    print("get file list time {:.5f} secs".format(elapsed_get_files))
    print("="*80)

    predictor = tf.contrib.predictor.from_saved_model(saved_model_dir)
    after_load_model = time.time()

    elapsed_load_model = after_load_model - after_get_files
    print("save_model load time {:.5f} secs".format(elapsed_load_model))
    print("="*80)

    for image_path in file_list:

        before_open_image = time.time()
        image = open(image_path, "rb").read()
        after_open_image = time.time()

        result = predictor({"image":[image], "key": [image_path]})
        after_predict_image = time.time()

        result = result

        elapsed_time_open_image = after_open_image - before_open_image
        elapsed_predict_image = after_predict_image - after_open_image
        file_name = os.path.basename(str(result['key'][0].decode("utf-8")))
        predicted_label = result['label'][0].decode("utf-8")
        max_score = max(result['score'][0])

        msg = "open:{:.5f} secs".format(elapsed_time_open_image) + "/"
        msg += "predict:{:.5f} secs".format(elapsed_predict_image) + " => "
        msg += "key:" + file_name + ","
        msg += "label:" + predicted_label + ","
        msg += "score:" + "{:.5f}".format(max_score)
        print(msg)

[/js]