diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
index ea4f4d3..e494129 100644
--- a/.pre-commit-config.yaml
+++ b/.pre-commit-config.yaml
@@ -4,7 +4,7 @@ repos:
     hooks:
       - id: flake8
   - repo: https://github.com/PyCQA/isort
-    rev: 5.10.1
+    rev: 5.11.5
     hooks:
       - id: isort
   - repo: https://github.com/pre-commit/mirrors-yapf
diff --git a/mmflow/datasets/AutoFlow.py b/mmflow/datasets/AutoFlow.py
new file mode 100644
index 0000000..6c33018
--- /dev/null
+++ b/mmflow/datasets/AutoFlow.py
@@ -0,0 +1,53 @@
+# Copyright (c) OpenMMLab. All rights reserved.
+import os
+import os.path as osp
+from typing import Any
+
+from .base_dataset import BaseDataset
+from .builder import DATASETS
+
+
+@DATASETS.register_module()
+class AutoFlow(BaseDataset):
+    """AutoFlow dataset."""
+
+    def __init__(self, *args: Any, **kwargs: Any):
+        super().__init__(*args, **kwargs)
+
+    def load_data_info(self) -> None:
+        """load data information."""
+
+        self.subset_dir = 'test' if self.test_mode else 'train'
+
+        self.data_root = osp.join(self.data_root, self.subset_dir)
+        self.img1_dir = self.data_root
+        self.img2_dir = self.data_root
+        self.flow_root = self.data_root
+
+        self.img_suffix = '.png'
+        self.flow_suffix = '.flo'
+
+        self.all_scene = os.listdir(self.img1_dir)
+
+        img1_filenames = []
+        img2_filenames = []
+        flow_filenames = []
+
+        for s in self.all_scene:
+            file_dir = os.listdir(osp.join(self.img1_dir, s))
+            for i in file_dir:
+                img_file_dir = osp.join(self.img1_dir, s, i)
+                flow_file_dir = osp.join(self.flow_root, s, i)
+
+                flow_filenames_ = self.get_data_filename(
+                    flow_file_dir, self.flow_suffix)
+                img_filenames_ = self.get_data_filename(
+                    img_file_dir, self.img_suffix)
+
+                flow_filenames.append(flow_filenames_[0])
+                img1_filenames.append(img_filenames_[0])
+                img2_filenames.append(img_filenames_[1])
+
+        # img1_filenames, img2_filenames = self._revise_dir(flow_filenames)
+        self.load_img_info(self.data_infos, img1_filenames, img2_filenames)
+        self.load_ann_info(self.data_infos, flow_filenames, 'filename_flow')
diff --git a/mmflow/datasets/CrowdFlow.py b/mmflow/datasets/CrowdFlow.py
new file mode 100644
index 0000000..1492cf3
--- /dev/null
+++ b/mmflow/datasets/CrowdFlow.py
@@ -0,0 +1,52 @@
+# Copyright (c) OpenMMLab. All rights reserved.
+import os
+import os.path as osp
+from typing import Any
+
+from .base_dataset import BaseDataset
+from .builder import DATASETS
+
+
+@DATASETS.register_module()
+class CrowdFlow(BaseDataset):
+    """CrowdFlow dataset."""
+
+    def __init__(self, *args: Any, **kwargs: Any):
+        super().__init__(*args, **kwargs)
+
+    def load_data_info(self) -> None:
+        """load data information."""
+
+        self.subset_dir = 'test' if self.test_mode else 'train'
+
+        self.data_root = osp.join(self.data_root, self.subset_dir)
+        self.img1_dir = osp.join(self.data_root, 'images')
+        self.img2_dir = osp.join(self.data_root, 'images')
+        self.flow_root = osp.join(self.data_root, 'gt_flow')
+
+        self.img_suffix = '.png'
+        self.flow_suffix = '.flo'
+
+        self.all_scene = os.listdir(self.img1_dir)
+
+        img1_filenames = []
+        img2_filenames = []
+        flow_filenames = []
+
+        for s in self.all_scene:
+            img_file_dir = osp.join(self.img1_dir, s)
+            flow_file_dir = osp.join(self.flow_root, s)
+
+            flow_filenames_ = self.get_data_filename(flow_file_dir,
+                                                     self.flow_suffix)
+            img_filenames_ = self.get_data_filename(img_file_dir,
+                                                    self.img_suffix)
+            flow_num = len(flow_filenames_)
+            for i in range(flow_num):
+                flow_filenames += [flow_filenames_[i]]
+                img1_filenames += [img_filenames_[i]]
+                img2_filenames += [img_filenames_[i + 1]]
+
+        # img1_filenames, img2_filenames = self._revise_dir(flow_filenames)
+        self.load_img_info(self.data_infos, img1_filenames, img2_filenames)
+        self.load_ann_info(self.data_infos, flow_filenames, 'filename_flow')
diff --git a/mmflow/datasets/__init__.py b/mmflow/datasets/__init__.py
index 5631e21..4a1c53e 100644
--- a/mmflow/datasets/__init__.py
+++ b/mmflow/datasets/__init__.py
@@ -1,6 +1,8 @@
 # Copyright (c) OpenMMLab. All rights reserved.
+from .AutoFlow import AutoFlow
 from .builder import DATASETS, PIPELINES, build_dataloader, build_dataset
 from .chairssdhom import ChairsSDHom
+from .CrowdFlow import CrowdFlow
 from .dataset_wrappers import ConcatDataset, RepeatDataset
 from .flyingchairs import FlyingChairs
 from .flyingchairsocc import FlyingChairsOcc
@@ -33,5 +35,5 @@
     'read_flow_kitti', 'GaussianNoise', 'RandomTranslate', 'Compose',
     'InputPad', 'FlyingThings3DSubset', 'FlyingThings3D', 'Sintel',
     'KITTI2012', 'KITTI2015', 'ChairsSDHom', 'HD1K', 'FlyingChairsOcc',
-    'render_color_wheel'
+    'CrowdFlow', 'AutoFlow', 'render_color_wheel'
 ]