bug fix

@@ -115,11 +115,18 @@ class InitModel(object):
                 
                 
         if compare_type != 0 and self.abnormal_face:
         if compare_type != 0 and self.abnormal_face:
             new_life_face_bboxes = []
             new_life_face_bboxes = []
-            print('1'*100)
+            abnormal_bboxes = []
+            is_abnormal = False
             for box_idx, face_bbox in enumerate(life_face_bboxes):            
             for box_idx, face_bbox in enumerate(life_face_bboxes):            
                 abnormal_pre = self.abnormal_recognizer.reg(life_rgb_image, face_bbox)
                 abnormal_pre = self.abnormal_recognizer.reg(life_rgb_image, face_bbox)
                 if abnormal_pre == 1:
                 if abnormal_pre == 1:
                     new_life_face_bboxes.append(face_bbox)
                     new_life_face_bboxes.append(face_bbox)
+                else:
+                    abnormal_bboxes.append([face_bbox])
+                    is_abnormal = True
+
+            if is_abnormal:
+                result_dict['abnormal_bboxes'] = abnormal_bboxes
  
  
             life_face_bboxes = new_life_face_bboxes      
             life_face_bboxes = new_life_face_bboxes      
 
 
@@ -210,7 +217,10 @@ if __name__ == '__main__':
     pre_pos = 0.01
     pre_pos = 0.01
     pre_all = 0.01
     pre_all = 0.01
     positive_num = 0.01
     positive_num = 0.01
-    for image_list_txt_line in image_list_txt_lines:
+    total_cost = 0
+    for image_index, image_list_txt_line in enumerate(image_list_txt_lines):
+        if image_index >= 10:
+            break
         arr = image_list_txt_line.strip().split(' ')
         arr = image_list_txt_line.strip().split(' ')
         label = arr[-1]
         label = arr[-1]
     
     
@@ -230,15 +240,24 @@ if __name__ == '__main__':
         life_image_dir = life_image_name_arr[-2]
         life_image_dir = life_image_name_arr[-2]
         life_image_path = os.path.join(image_dir, life_image_dir, life_image_name)
         life_image_path = os.path.join(image_dir, life_image_dir, life_image_name)
 
 
+        id_image_path = r'input/0003/0.jpg'
+        life_image_path = r'input/0003/1.jpg'
+
         id_image = cv2.imread(id_image_path)
         id_image = cv2.imread(id_image_path)
         life_image = cv2.imread(life_image_path)
         life_image = cv2.imread(life_image_path)
     
     
         r, id_image_str = cv2.imencode('.jpg', id_image)
         r, id_image_str = cv2.imencode('.jpg', id_image)
         id_base64_image = base64.b64encode(id_image_str)
         id_base64_image = base64.b64encode(id_image_str)
 
 
+        with open('base64_0.txt', 'w') as f:
+            f.write(str(id_base64_image))
+
         r, life_image_str = cv2.imencode('.jpg', life_image)
         r, life_image_str = cv2.imencode('.jpg', life_image)
         life_base64_image = base64.b64encode(life_image_str)
         life_base64_image = base64.b64encode(life_image_str)
         
         
+        with open('base64_1.txt', 'w') as f:
+            f.write(str(life_base64_image))
+    
         st = time.time()
         st = time.time()
         result_dict = {}
         result_dict = {}
         if format_type == 0:
         if format_type == 0:
@@ -247,9 +266,9 @@ if __name__ == '__main__':
             result_dict = main(id_image, life_image, compare_type)
             result_dict = main(id_image, life_image, compare_type)
         print(result_dict)
         print(result_dict)
         et = time.time()
         et = time.time()
+        total_cost = total_cost + (et - st)
         print('total time cost:{}'.format(round((et-st), 2)))
         print('total time cost:{}'.format(round((et-st), 2)))
         
         
-        print(label) 
         res = result_dict['match']
         res = result_dict['match']
         if res:
         if res:
             if label == '1':
             if label == '1':
@@ -260,6 +279,7 @@ if __name__ == '__main__':
             if label == '0':
             if label == '0':
                 hit += 1
                 hit += 1
     
     
+    print('average time cost:{}'.format(total_cost / 10))
     print('precision:{}'.format(hit_pos/pre_pos))
     print('precision:{}'.format(hit_pos/pre_pos))
     print('recall:{}'.format(hit_pos/positive_num))
     print('recall:{}'.format(hit_pos/positive_num))
     print('accuracy:{}'.format(hit/pre_all))            
     print('accuracy:{}'.format(hit/pre_all))            

@@ -26,8 +26,8 @@ class Abnormal_Face(object):
         self.interpreter.runSession(self.session)
         self.interpreter.runSession(self.session)
         output_tensor = self.interpreter.getSessionOutput(self.session)
         output_tensor = self.interpreter.getSessionOutput(self.session)
         output_data = output_tensor.getData()
         output_data = output_tensor.getData()
-        output_data = output_data[0]
-        output_data = output_data.tolist()
+        # output_data = output_data[0]
+        # output_data = output_data.tolist()
         predict_res = output_data.index(max(output_data))
         predict_res = output_data.index(max(output_data))
        
        
         return predict_res
         return predict_res

@@ -11,12 +11,47 @@ class Face_Recognizer(object):
         self.reg_input_tensor = self.reg_interpreter.getSessionInput(self.reg_session)
         self.reg_input_tensor = self.reg_interpreter.getSessionInput(self.reg_session)
         logging.info('******** Success Init Face ID  ********')
         logging.info('******** Success Init Face ID  ********')
 
 
+#     def recognize(self, imgs):
+#         
+#         new_imgs = []
+#         for i in range(len(imgs)):
+#             img = imgs[i] 
+#             img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+#             img = (img / 255. - 0.5) / 0.5
+#             img = img.transpose(2, 0, 1)
+#             img = np.expand_dims(img, axis=0)
+#             
+#             new_imgs.append(img) 
+#    
+#         new_imgs = np.array(new_imgs)
+#         new_imgs = new_imgs.astype(np.float32)
+#             
+#         self.reg_interpreter.resizeTensor(self.reg_input_tensor, (len(imgs), 3, 112, 112))
+#         self.reg_interpreter.resizeSession(self.reg_session)               
+#         
+#         input_tensor = MNN.Tensor((len(imgs), 3, 112, 112), MNN.Halide_Type_Float, new_imgs, MNN.Tensor_DimensionType_Caffe)
+#         self.reg_input_tensor.copyFrom(input_tensor)
+#         self.reg_interpreter.runSession(self.reg_session)
+#         output_tensor = self.reg_interpreter.getSessionOutput(self.reg_session, 'output0')
+#         output = output_tensor.getData()
+#         
+#         # feats.append(output[0])
+#         # feats.append(output)
+#         
+#         embeddings = list()
+#         num = int(len(output)/512)
+#         for i in range(num):
+#             if i < num:
+#                 embeddings.append(output[i*512:(i+1)*512])
+# 
+#         feats_np = np.array(embeddings)
+#         return feats_np
+
     def recognize(self, imgs):
     def recognize(self, imgs):
         
         
         feats = []
         feats = []
         for i in range(len(imgs)):
         for i in range(len(imgs)):
             img = imgs[i] 
             img = imgs[i] 
-
             img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
             img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
             img = img.astype(np.float)
             img = img.astype(np.float)
             img = (img / 255. - 0.5) / 0.5
             img = (img / 255. - 0.5) / 0.5
@@ -32,6 +67,5 @@ class Face_Recognizer(object):
             
             
             # feats.append(output[0])
             # feats.append(output[0])
             feats.append(output)
             feats.append(output)
-
         feats_np = np.array(feats)
         feats_np = np.array(feats)
         return feats_np
         return feats_np

-{"code": 2000, "id_face_bboxes": [[141, 94, 361, 380], [42, 133, 56, 156]], "id_face_landmarks": [[[196, 200], [290, 204], [235, 246], [202, 304], [277, 308]], [[51, 142], [53, 142], [54, 147], [50, 151], [52, 152]]], "life_face_bboxes": [[832, 187, 1032, 383]], "life_face_landmarks": [[[890, 261], [971, 258], [937, 294], [911, 339], [968, 336]]], "id_index": 0, "life_index": 0, "sim": 0.3722387966977346, "match": true}
\ No newline at end of file
+{"code": 2000, "id_face_bboxes": [[572, 142, 713, 305]], "id_face_landmarks": [[[610, 206], [681, 207], [645, 237], [617, 268], [666, 269]]], "life_face_bboxes": [[476, 125, 606, 292]], "life_face_landmarks": [[[509, 191], [572, 192], [537, 219], [514, 253], [560, 253]]], "match": false}
\ No newline at end of file