diff --git a/torch_em/data/datasets/deepbacs.py b/torch_em/data/datasets/deepbacs.py
index 32bd582b..4149128e 100644
--- a/torch_em/data/datasets/deepbacs.py
+++ b/torch_em/data/datasets/deepbacs.py
@@ -1,4 +1,7 @@
 import os
+import shutil
+import numpy as np
+from glob import glob
 
 import torch_em
 from . import util
@@ -25,18 +28,61 @@ def _require_deebacs_dataset(path, bac_type, download):
         util.download_source(zip_path, URLS[bac_type], download, checksum=CHECKSUMS[bac_type])
     util.unzip(zip_path, os.path.join(path, bac_type))
 
+    # let's get a val split for the expected bacteria type
+    _assort_val_set(path, bac_type)
+
+
+def _assort_val_set(path, bac_type):
+    image_paths = glob(os.path.join(path, bac_type, "training", "source", "*"))
+    image_paths = [os.path.split(_path)[-1] for _path in image_paths]
+
+    val_partition = 0.2
+    # let's get a balanced set of bacterias, if bac_type is mixed
+    if bac_type == "mixed":
+        _sort_1, _sort_2, _sort_3 = [], [], []
+        for _path in image_paths:
+            if _path.startswith("JE2"):
+                _sort_1.append(_path)
+            elif _path.startswith("pos"):
+                _sort_2.append(_path)
+            elif _path.startswith("train_"):
+                _sort_3.append(_path)
+
+        val_image_paths = [
+            *np.random.choice(_sort_1, size=int(val_partition * len(_sort_1)), replace=False),
+            *np.random.choice(_sort_2, size=int(val_partition * len(_sort_2)), replace=False),
+            *np.random.choice(_sort_3, size=int(val_partition * len(_sort_3)), replace=False)
+        ]
+    else:
+        val_image_paths = np.random.choice(image_paths, size=int(val_partition * len(image_paths)), replace=False)
+
+    val_image_dir = os.path.join(path, bac_type, "val", "source")
+    val_label_dir = os.path.join(path, bac_type, "val", "target")
+    os.makedirs(val_image_dir, exist_ok=True)
+    os.makedirs(val_label_dir, exist_ok=True)
+
+    for sample_id in val_image_paths:
+        src_val_image_path = os.path.join(path, bac_type, "training", "source", sample_id)
+        dst_val_image_path = os.path.join(val_image_dir, sample_id)
+        shutil.move(src_val_image_path, dst_val_image_path)
+
+        src_val_label_path = os.path.join(path, bac_type, "training", "target", sample_id)
+        dst_val_label_path = os.path.join(val_label_dir, sample_id)
+        shutil.move(src_val_label_path, dst_val_label_path)
+
 
 def _get_paths(path, bac_type, split):
     # the bacteria types other than mixed are a bit more complicated so we don't have the dataloaders for them yet
     # mixed is the combination of all other types
+    if split == "train":
+        dir_choice = "training"
+    else:
+        dir_choice = split
+
     if bac_type != "mixed":
         raise NotImplementedError(f"Currently only the bacteria type 'mixed' is supported, not {bac_type}")
-    image_folder = os.path.join(
-        path, bac_type, "training" if split == "train" else "test", "source"
-    )
-    label_folder = os.path.join(
-        path, bac_type, "training" if split == "train" else "test", "target"
-    )
+    image_folder = os.path.join(path, bac_type, dir_choice, "source")
+    label_folder = os.path.join(path, bac_type, dir_choice, "target")
     return image_folder, label_folder
 
 
@@ -48,7 +94,7 @@ def get_deepbacs_dataset(
     This dataset is from the publication https://doi.org/10.1038/s42003-022-03634-z.
     Please cite it if you use this dataset for a publication.
     """
-    assert split in ("train", "test")
+    assert split in ("train", "val", "test")
     bac_types = list(URLS.keys())
     assert bac_type in bac_types, f"{bac_type} is not in expected bacteria types: {bac_types}"