add func to create the folder for save snapshot

notebook/face_predict.py

@@ -10,6 +10,7 @@ from skimage.transform import resize
 from sklearn.preprocessing import LabelEncoder
 from keras.models import model_from_json
 import numpy as np
+import os
 
 
 # In[2]:
@@ -34,12 +35,25 @@ print("Facenet預測模型載入完成")
 
 
 # 載入SVM分類器
-clf=joblib.load('../model/20190622181256/20190622181256.pkl')
+clf=joblib.load('../model/20190624135811/20190624135811.pkl')
 # 載入LabelEncoder
 le=LabelEncoder()
-le.classes_ =np.load('../model/20190622181256/classes.npy')
+le.classes_ =np.load('../model/20190624135811/classes.npy')
 print("SVM分類器載入完成")
 
+# 建立儲存影像位置
+with open('../model/20190624135811//labels.txt') as f:
+    lines = f.readlines()
+    
+for i in range(len(lines)):
+    name=lines[i].split('\n')[0]
+    save_path=os.path.abspath('../save/')
+    name_path=os.path.join(save_path,name)
+    if(not os.path.isdir(name_path)):
+        os.mkdir(name_path)
+        os.mkdir(os.path.join(name_path,'login_'))
+        os.mkdir(os.path.join(name_path,'logout_'))
+        os.mkdir(os.path.join(name_path,'error_'))
 # In[5]:
 
 
@@ -107,16 +121,25 @@ def infer(le, clf, img):
     
     embs = calc_embs(face_cropped(img,faces,10))
 #     pred = le.inverse_transform(clf.predict(embs))
-    pred=get_labels(le,clf,embs)
-    return [faces,pred]
+#     pred=get_labels(le,clf,embs)
+    results,confidences=get_labelsNconfidence(le,clf,embs)
+    return [faces,results,confidences]
 # Labels 解析
 def get_labels(le,clf,embs):
     socres=clf.predict_proba(embs)
     results=[]
     for s in socres:
-        print(s)
         if(s[s.argmax()]>0.5):
             results.append(le.inverse_transform([s.argmax()])[0])
         else:
             results.append('Unknow')
-    return results
\ No newline at end of file
+    return results
+# Labels 解析回傳結果與分數
+def get_labelsNconfidence(le,clf,embs):
+    socres=clf.predict_proba(embs)
+    results=[]
+    confidences=[]
+    for s in socres:
+        results.append(le.inverse_transform([s.argmax()])[0])
+        confidences.append(s[s.argmax()])
+    return results,confidences
\ No newline at end of file