add load_test_data(test_alpha)

matting.py

@@ -113,9 +113,10 @@ def preprocessing_single(comp_RGB, alpha, BG, FG, batch_RGB_paths,i,image_size=3
 
  if crop_size == 620:
  h_start_index = i_UR_center[0] - 309
- w_start_index = i_UR_center[1] - 309
+ #boundary security
  if h_start_index<0:
  h_start_index = 0
+ w_start_index = i_UR_center[1] - 309
  if w_start_index<0:
  w_start_index = 0
  tmp = train_pre[h_start_index:h_start_index+620, w_start_index:w_start_index+620, :]
@@ -129,4 +130,26 @@ def preprocessing_single(comp_RGB, alpha, BG, FG, batch_RGB_paths,i,image_size=3
  tmp1[:,:,8:] = misc.imresize(tmp[:,:,8:],[image_size,image_size,3])
  train_data = tmp1
  train_data = train_data.astype(np.float32)
- return train_data
+ return train_data
+
+def load_test_data(test_alpha):
+ rgb_path = os.path.join(test_alpha,'rgb')
+ trimap_path = os.path.join(test_alpha,'trimap')
+ alpha_path = os.path.join(test_alpha,'alpha') 
+ images = os.listdir(trimap_path)
+ test_num = len(images)
+ all_shape = []
+ rgb_batch = []
+ tri_batch = []
+ alp_batch = []
+ for i in range(test_num):
+ rgb = misc.imread(os.path.join(rgb_path,images[i]))
+ trimap = misc.imread(os.path.join(trimap_path,images[i]),'L')
+ alpha = misc.imread(os.path.join(alpha_path,images[i]),'L')
+ all_shape.append(trimap.shape)
+ rgb_batch.append(misc.imresize(rgb,[320,320,3]))
+ trimap = misc.imresize(trimap,[320,320],interp = 'nearest')
+ tri_batch.append(np.expand_dims(trimap,2))
+ alp_batch.append(np.expand_dims(alpha,2))
+ return np.array(rgb_batch),np.array(tri_batch),np.array(alp_batch),all_shape,images
+