fix code style problem

tools/train_lanenet.py

@@ -378,17 +378,15 @@ def train_lanenet(dataset_dir, weights_path=None, net_flag='vgg'):
 
     # Set tf model save path
     model_save_dir = 'model/tusimple_lanenet_{:s}'.format(net_flag)
-    # if not ops.exists(model_save_dir):
-    #     os.makedirs(model_save_dir)
+    os.makedirs(model_save_dir, exist_ok=True)
     train_start_time = time.strftime('%Y-%m-%d-%H-%M-%S', time.localtime(time.time()))
     model_name = 'tusimple_lanenet_{:s}_{:s}.ckpt'.format(net_flag, str(train_start_time))
     model_save_path = ops.join(model_save_dir, model_name)
     saver = tf.train.Saver()
 
     # Set tf summary save path
     tboard_save_path = 'tboard/tusimple_lanenet_{:s}'.format(net_flag)
-    if not ops.exists(tboard_save_path):
-        os.makedirs(tboard_save_path)
+    os.makedirs(tboard_save_path, exist_ok=True)
 
     # Set sess configuration
     sess_config = tf.ConfigProto(allow_soft_placement=True)
@@ -425,9 +423,10 @@ def train_lanenet(dataset_dir, weights_path=None, net_flag='vgg'):
             # training part
             t_start = time.time()
 
-            _, train_c, train_accuracy_figure, train_fn_figure, train_fp_figure, lr, train_summary, train_binary_loss, \
-            train_instance_loss, train_embeddings, train_binary_seg_imgs, train_gt_imgs, \
-            train_binary_gt_labels, train_instance_gt_labels = \
+            _, train_c, train_accuracy_figure, train_fn_figure, train_fp_figure, \
+                lr, train_summary, train_binary_loss, \
+                train_instance_loss, train_embeddings, train_binary_seg_imgs, train_gt_imgs, \
+                train_binary_gt_labels, train_instance_gt_labels = \
                 sess.run([optimizer, train_total_loss, train_accuracy, train_fn, train_fp,
                           learning_rate, train_merge_summary_op, train_binary_seg_loss,
                           train_disc_loss, train_pix_embedding, train_prediction,
@@ -456,9 +455,10 @@ def train_lanenet(dataset_dir, weights_path=None, net_flag='vgg'):
                 train_cost_time_mean.clear()
 
             # validation part
-            val_c, val_accuracy_figure, val_fn_figure, val_fp_figure, val_summary, val_binary_loss, \
-            val_instance_loss, val_embeddings, val_binary_seg_imgs, val_gt_imgs, \
-            val_binary_gt_labels, val_instance_gt_labels = \
+            val_c, val_accuracy_figure, val_fn_figure, val_fp_figure, \
+                val_summary, val_binary_loss, val_instance_loss, \
+                val_embeddings, val_binary_seg_imgs, val_gt_imgs, \
+                val_binary_gt_labels, val_instance_gt_labels = \
                 sess.run([val_total_loss, val_accuracy, val_fn, val_fp,
                           val_merge_summary_op, val_binary_seg_loss,
                           val_disc_loss, val_pix_embedding, val_prediction,