準備工作

首先，需要在Microsoft的主頁（https://azure.microsoft.com/zh-cn/）註冊一個賬號，然後進入門戶去創建新資源，選擇AI+Cognitive Services中的人臉API，填寫相關信息就可以了。

微軟Azure的免費API是限制每分鐘的訪問量與月訪問量的，其他功能倒是沒什麽區別。但是之前創建這個訂閱是不需要綁定信用卡就可以獲取API Key和API Secret的，後來再創建的時候發現必須要綁定Visa信用卡才可以了（？）

總之到這裏，我們拿到了API Key,API Secret和URL。

然後準備好本地待識別情緒的圖片/相片。

代碼

介紹下所使用的第三方庫

——httplib是一個相對底層的http請求模塊

——urllib是接受URL請求的相關模塊

——json （Emmmmmm……我也不知道該怎麽解釋這個）

——xlwt是對excel進行寫入操作的一個庫

——time是對時間進行處理的一個庫，以下代碼中其實就使用了sleep()和localtime()兩個函數，sleep()是用來讓程序暫停幾秒的，localtime()是格式化時間戳為本地的時間

——os是操作系統的相關功能的一個庫，例如用來處理文件和目錄之類的

  1 # coding:utf-8
  2 # version:python2.7.6
  3 # author:Ivy Wong
 

  4 
  5 # 導入相關模塊
  6 import httplib, urllib, json
  7 import xlwt, time, os
  8 
  9 
 10 
 11 
 12 # 使用micrsoft的api識別情緒
 13 def useapi(img):
 14     # 定義html的header，這裏Content-type決定了body中的類型，是URL還是文件類型的，這裏的Json支持URL模式
 15     headers = {
 16         ‘Content-Type‘: ‘application/octet-stream‘,
 17         ‘
 
Ocp-Apim-Subscription-Key‘: subscription_key,
 18     }
 19 
 20     # 定義返回的內容，包括FaceId，年齡、性別等等
 21     params = urllib.urlencode({
 22         ‘returnFaceId‘: ‘true‘,
 23         ‘returnFaceLandmarks‘: ‘false‘,
 24         ‘returnFaceAttributes‘: ‘age,gender,smile,glasses,emotion‘,
 25     })
 26 
 27     # Call Face API，進行人臉識別
 28     try:
 29         # Execute the REST API call and get the response.
 30         conn = httplib.HTTPSConnection(‘api.cognitive.azure.cn‘)
 31         conn.request("POST", "/face/v1.0/detect?%s" % params, img, headers)
 32         response = conn.getresponse()
 33         data = response.read()
 34 
 35         # ‘data‘ contains the JSON data. The following formats the JSON data for display.
 36         parsed = json.loads(data)
 37         conn.close()
 38 
 39     except Exception as e:
 40         print("[Errno {0}] {1}".format(e.errno, e.strerror))
 41 
 42     return parsed
 43 
 44 
 45 # 將json字典寫入excel
 46 # 變量用來循環時控制寫入單元格，感覺這種方式有點傻，但暫時想不到優化方法
 47 def writeexcel(img, worksheet, row, files_name, path,file_num):
 48     parsed = useapi(img)
 49     if not parsed:
 50         print ‘This picture do not have any face‘
 51     elif ‘error‘ in parsed:
 52         print parsed[‘error‘][‘message‘]
 53         # 如果是達到每分鐘的限制，就先暫停一分鐘 
 54         if parsed[‘error‘][‘message‘]==‘Rate limit is exceeded. Try again later.‘:
 55             print ‘The file number is ‘ + str(file_num)
 56             print ‘The program will be asleep for 60s‘
 57             time.sleep(60)
 58             print ‘Now, the program go back to work!‘
 59             writeexcel(img,worksheet,row,files_name,path,file_num)
 60     else:
 61         for list_item in parsed:
 62             # 寫入文件名
 63             filename, extension=os.path.splitext(files_name)
 64             worksheet.write(row, 0, filename)
 65 
 66             # 寫入時間戳
 67             daystamp, timestamp, hourstamp = gettimestamp(path)
 68             worksheet.write(row, 1, label=daystamp)
 69             worksheet.write(row, 2, label=timestamp)
 70             worksheet.write(row,3,hourstamp)
 71 
 72             # 寫入api返回的數據
 73             emotion = []
 74             for key1, value1 in list_item.items():
 75                 if key1 == ‘faceAttributes‘:
 76                     for key2, value2 in value1.items():
 77                         if key2 == ‘age‘:
 78                             worksheet.write(row, 5, value2)
 79                         elif key2 == ‘emotion‘:
 80                             for key3, value3 in value2.items():
 81                                 if key3 == ‘anger‘:
 82                                     worksheet.write(row, 8, value3)
 83                                     emotion.append(value3)
 84                                 elif key3 == ‘contempt‘:
 85                                     worksheet.write(row, 9, value3)
 86                                     emotion.append(value3)
 87                                 elif key3 == ‘disgust‘:
 88                                     worksheet.write(row, 10, value3)
 89                                     emotion.append(value3)
 90                                 elif key3 == ‘fear‘:
 91                                     worksheet.write(row, 11, value3)
 92                                     emotion.append(value3)
 93                                 elif key3 == ‘happiness‘:
 94                                     worksheet.write(row, 12, value3)
 95                                     emotion.append(value3)
 96                                 elif key3 == ‘neutral‘:
 97                                     worksheet.write(row, 13, value3)
 98                                     emotion.append(value3)
 99                                 elif key3 == ‘sadness‘:
100                                     worksheet.write(row, 14, value3)
101                                     emotion.append(value3)
102                                 else:
103                                     worksheet.write(row, 15, value3)
104                                     emotion.append(value3)
105                         elif key2 == ‘gender‘:
106                             worksheet.write(row, 6, value2)
107                         elif key2 == ‘glasses‘:
108                             worksheet.write(row, 7, value2)
109                         else:
110                             pass
111                 elif key1 == ‘faceId‘:
112                     worksheet.write(row, 4, value1)
113                 else:
114                     pass
115             worksheet.write(row, 16, emotion.index(max(emotion)))
116             # 寫入概率最大的情緒，0-sadness,1-neutral,2-contempt，3-disgust，4-anger，5-surprise，6-fear，7-happiness
117             row += 1
118             print ‘Success! The pic ‘ + str(files_name) + ‘ was detected!‘
119     return row, worksheet
120 
121 
122 # 獲取時間戳
123 def gettimestamp(path):
124     statinfo = os.stat(path)
125     timeinfo = time.localtime(statinfo.st_ctime)
126     daystamp = str(timeinfo.tm_year) + ‘-‘ + str(timeinfo.tm_mon) + ‘-‘ + str(timeinfo.tm_mday)
127     timestamp = str(timeinfo.tm_hour) + ‘:‘ + str(timeinfo.tm_min) + ‘:‘ + str(timeinfo.tm_sec)
128     hourstamp = timeinfo.tm_hour+timeinfo.tm_min/60.0+timeinfo.tm_sec/3600.0
129     return daystamp, timestamp, hourstamp
130 
131 
132 subscription_key = ‘your_key‘
133 uri_base = ‘https://api.cognitive.azure.cn/face/v1.0‘
134 path = r‘圖片文件夾路徑‘
135 # 註意：由於我是對同一文件夾下的多個文件夾中的圖片進行識別，所以這個path是圖片所在文件夾的上一級文件夾。文件夾名盡量使用英文與數字，不然可能因為編碼問題報錯
136 # 創建excel
137 workbook = xlwt.Workbook(encoding=‘utf-8‘)
138 
139 
140 
141 for root, dirs, files in os.walk(path, topdown=False):
142     for folder in dirs:
143 
144         error_num = 0
145         error_list = []
146         print ‘Let us start dealing with folder ‘ + folder
147 
148         # 創建一個新的sheet
149         worksheet = workbook.add_sheet(folder)
150         # 設置表頭
151         title = [‘PhotoID‘, ‘daystamp‘, ‘timestamp‘, ‘hourstamp‘,‘faceID‘, ‘age‘, ‘gender‘,
152                  ‘glasses‘, ‘anger‘, ‘contempt‘,‘disgust‘, ‘fear‘, ‘happiness‘,
153                  ‘neutral‘, ‘sadness‘, ‘surprise‘,‘emotion‘]
154         for col in range(len(title)):
155             worksheet.write(0, col, title[col])
156 
157         # 遍歷每個folder裏的圖片
158         row = 1
159         file_num = 1
160         for root2, dirs2, files2 in os.walk(path + ‘\\‘ + folder):
161             for files_name in files2:
162                 try:
163                     path2 = path + ‘\\‘ + folder + ‘\\‘ + files_name
164                     print ‘Now, the program is going to deal with ‘ + folder + ‘ pic‘ + str(files_name)
165                     img = open(os.path.expanduser(path2), ‘rb‘)
166                     row, worksheet = writeexcel(img, worksheet, row, files_name, path2,file_num)
167                     file_num += 1
168                 except Exception as e:
169                     print e
170 
171 
172         print ‘The current folder is done.‘
173         print error_num,error_list
174 
175 # 保存文件
176 workbook.save(‘detectface.xls‘)
177 print ‘The program is done!‘    

成果

最後生成的excel大概是這個樣子。

其中emotion就是概率最大的情緒，0-sadness,1-neutral,2-contempt，3-disgust，4-anger，5-surprise，6-fear，7-happiness。face++返回的有7種情緒，而azure返回的有8種情緒。

【Python】使用Microsoft Azure的人臉API進行本地圖片情緒識別並存入excel