diff --git a/README.md b/README.md
index 752efb0..4729d8b 100644
--- a/README.md
+++ b/README.md
@@ -31,7 +31,6 @@ Then update the train/val/test list paths in the config files.
 #### [Update] We have uploaded the lists we use in our paper.
 + The train/val lists for COCO contain 82081 and 40137 images respectively. They are generated by the default scripts provided by COCO.
 + The train/val lists for PASCAL5i contain 5953 and 1449 images respectively. The train list should be **voc_sbd_merge_noduplicate.txt** and the val list is the original val list of pascal voc (**val_fixed.txt**).
-+ We do not use the augmented images of voc as the training data.
 
 ##### To get voc_sbd_merge_noduplicate.txt:
 + we first merge the original VOC (voc_original_train.txt) and SBD ([**sbd_data.txt**](http://home.bharathh.info/pubs/codes/SBD/train_noval.txt)) training data.