Update load_meta_cnn.py

emotion_classifier_tensorflow_version/CNN/load_meta_cnn.py

@@ -22,7 +22,7 @@
 channel = 1
 default_height = 48
 default_width = 48
-confusion_matrix = True
+confusion_matrix = False
 use_advanced_method = True
 emotion_labels = ['angry', 'disgust:', 'fear', 'happy', 'sad', 'surprise', 'neutral']
 num_class = len(emotion_labels)
@@ -41,14 +41,7 @@
 logits = graph.get_tensor_by_name('project/output/logits:0')
 
 
-def produce_result(images_):
- images_ = np.multiply(np.array(images_), 1./255)
- pred_logits_ = sess.run(tf.nn.softmax(logits), {x_input: images_, dropout: 1.0})
- return pred_logits_
-
-
-def produce_advanced_result(images_):
- images_ = np.multiply(np.array(images_), 1. / 255)
+def advance_image(images_):
  rsz_img = []
  rsz_imgs = []
  for image_ in images_:
@@ -64,6 +57,15 @@ def produce_advanced_result(images_):
  rsz_img.append(np.reshape(cv2.resize(cv2.flip(image_, 1), (default_height, default_width)),
  [default_height, default_width, channel]))
  rsz_imgs.append(np.array(rsz_img))
+ return rsz_imgs
+
+
+def produce_result(images_):
+ images_ = np.multiply(np.array(images_), 1. / 255)
+ if use_advanced_method:
+ rsz_imgs = advance_image(images_)
+ else:
+ rsz_imgs = [images_]
  pred_logits_ = []
  for rsz_img in rsz_imgs:
  pred_logits_.append(sess.run(tf.nn.softmax(logits), {x_input: rsz_img, dropout: 1.0}))
@@ -72,10 +74,7 @@ def produce_advanced_result(images_):
 
 def produce_results(images_):
  results = []
- if use_advanced_method:
- pred_logits_ = produce_advanced_result(images_)
- else:
- pred_logits_ = produce_result(images_)
+ pred_logits_ = produce_result(images_)
  pred_logits_list_ = np.array(np.reshape(np.argmax(pred_logits_, axis=1), [-1])).tolist()
  for num in range(num_class):
  results.append(pred_logits_list_.count(num))
@@ -102,10 +101,7 @@ def produce_confusion_matrix(images_list_, total_num_):
 def predict_emotion(image_):
  image_ = cv2.resize(image_, (default_height, default_width))
  image_ = np.reshape(image_, [-1, default_height, default_width, channel])
- if use_advanced_method:
- return produce_advanced_result(image_)[0]
- else:
- return produce_result(image_)[0]
+ return produce_result(image_)[0]
 
 
 def face_detect(image_path, casc_path_=casc_path):