diff --git a/ppdet/data/data_feed.py b/ppdet/data/data_feed.py
index d5a50696b6c17cff3007e3adf2450c34c895b4ba..e15b9da7ceb04e3ba34b6d5e870703648c0d8349 100644
--- a/ppdet/data/data_feed.py
+++ b/ppdet/data/data_feed.py
@@ -42,7 +42,7 @@ __all__ = [
 ]
 
 
-def create_reader(feed, max_iter=0):
+def create_reader(feed, max_iter=0, args_path=None):
     """
     Return iterable data reader.
 
@@ -52,11 +52,11 @@ def create_reader(feed, max_iter=0):
 
     # if `DATASET_DIR` does not exists, search ~/.paddle/dataset for a directory
     # named `DATASET_DIR` (e.g., coco, pascal), if not present either, download
-    if feed.dataset.dataset_dir:
+    dataset_home = args_path if args_path else feed.dataset.dataset_dir
+    if dataset_home:
         annotation = getattr(feed.dataset, 'annotation', None)
         image_dir = getattr(feed.dataset, 'image_dir', None)
-        dataset_dir = get_dataset_path(feed.dataset.dataset_dir,
-                                       annotation, image_dir)
+        dataset_dir = get_dataset_path(dataset_home, annotation, image_dir)
         if annotation:
             feed.dataset.annotation = os.path.join(dataset_dir, annotation)
         if image_dir:
diff --git a/tools/eval.py b/tools/eval.py
index 1e104b132eeb706aa08216505a0fb4f0bb4c906e..8ed450074b9003260fe7790c0eba149feed61fff 100644
--- a/tools/eval.py
+++ b/tools/eval.py
@@ -75,7 +75,7 @@ def main():
             fetches = model.eval(feed_vars)
     eval_prog = eval_prog.clone(True)
 
-    reader = create_reader(eval_feed)
+    reader = create_reader(eval_feed, args_path=FLAGS.dataset_dir)
     pyreader.decorate_sample_list_generator(reader, place)
 
     # compile program for multi-devices
@@ -114,8 +114,8 @@ def main():
     resolution = None
     if 'mask' in results[0]:
         resolution = model.mask_head.resolution
-    eval_results(results, eval_feed, cfg.metric, cfg.num_classes, 
-                 resolution, is_bbox_normalized, FLAGS.output_file)
+    eval_results(results, eval_feed, cfg.metric, cfg.num_classes, resolution,
+                 is_bbox_normalized, FLAGS.output_file)
 
 
 if __name__ == '__main__':
@@ -126,5 +126,11 @@ if __name__ == '__main__':
         default=None,
         type=str,
         help="Evaluation file name, default to bbox.json and mask.json.")
+    parser.add_argument(
+        "-d",
+        "--dataset_dir",
+        default=None,
+        type=str,
+        help="Dataset path, same as DataFeed.dataset.dataset_dir")
     FLAGS = parser.parse_args()
     main()
diff --git a/tools/train.py b/tools/train.py
index 906eb08a0060155375cf94d3ad53c0b7aa3640c6..7ffb204129cdab8c7bfd566c129d4cf5a6d5759d 100644
--- a/tools/train.py
+++ b/tools/train.py
@@ -105,7 +105,8 @@ def main():
             optimizer = optim_builder(lr)
             optimizer.minimize(loss)
 
-    train_reader = create_reader(train_feed, cfg.max_iters * devices_num)
+    train_reader = create_reader(train_feed, cfg.max_iters * devices_num,
+                                 FLAGS.dataset_dir)
     train_pyreader.decorate_sample_list_generator(train_reader, place)
 
     # parse train fetches
@@ -121,7 +122,7 @@ def main():
                 fetches = model.eval(feed_vars)
         eval_prog = eval_prog.clone(True)
 
-        eval_reader = create_reader(eval_feed)
+        eval_reader = create_reader(eval_feed, args_path=FLAGS.dataset_dir)
         eval_pyreader.decorate_sample_list_generator(eval_reader, place)
 
         # parse eval fetches
@@ -197,7 +198,7 @@ def main():
                 resolution = None
                 if 'mask' in results[0]:
                     resolution = model.mask_head.resolution
-                eval_results(results, eval_feed, cfg.metric, cfg.num_classes, 
+                eval_results(results, eval_feed, cfg.metric, cfg.num_classes,
                              resolution, is_bbox_normalized, FLAGS.output_file)
 
     train_pyreader.reset()
@@ -222,5 +223,11 @@ if __name__ == '__main__':
         default=None,
         type=str,
         help="Evaluation file name, default to bbox.json and mask.json.")
+    parser.add_argument(
+        "-d",
+        "--dataset_dir",
+        default=None,
+        type=str,
+        help="Dataset path, same as DataFeed.dataset.dataset_dir")
     FLAGS = parser.parse_args()
     main()