diff --git a/skluc/main/data/mldatasets/MnistDataset.py b/skluc/main/data/mldatasets/MnistDataset.py
index f2adb8f291706f86c97d0300ab572497ae64bb0d..0959025212893ce1eb4f641d5d1c79e312ce509c 100644
--- a/skluc/main/data/mldatasets/MnistDataset.py
+++ b/skluc/main/data/mldatasets/MnistDataset.py
@@ -55,11 +55,7 @@ class MnistDataset(ImageDataset):
 
     def read(self):
         """
-        Return a dict of data where, for each key is associated a (data, label) tuple.
-
-        The values of the tuple are np.ndarray.
-
-        :return: dict
+        set the _train and _test attribute of dataset
         """
         # todo add possibility to provide percentage for validation set instead of size
         self._train = LabeledData(
diff --git a/skluc/test/test_data/test_mldatasets/TestMnistDataset.py b/skluc/test/test_data/test_mldatasets/TestMnistDataset.py
new file mode 100644
index 0000000000000000000000000000000000000000..c2928edf8598fa6cfc205973bd01b7ea59ad4a09
--- /dev/null
+++ b/skluc/test/test_data/test_mldatasets/TestMnistDataset.py
@@ -0,0 +1,23 @@
+import os
+import unittest
+
+from skluc.main.data.mldatasets import MnistDataset
+
+
+class TestMnistDataset(unittest.TestCase):
+
+    def test_mnist(self):
+        mnist = MnistDataset()
+        mnist.load()
+        for name in mnist.l_filepaths:
+            self.assertTrue(os.path.exists(name))
+
+    def test_to_image(self):
+        mnist = MnistDataset()
+        mnist.load()
+        mnist.to_image()
+        self.assertTrue(mnist.train.data[0].shape == (28, 28, 1))
+
+
+if __name__ == "__main__":
+    unittest.main()