diff --git a/datasets.py b/datasets.py
index a5e59aa..d4753f5 100644
--- a/datasets.py
+++ b/datasets.py
@@ -42,6 +42,16 @@ def make_dataset_ots(root):
     return items
 
 
+def make_dataset_ohaze(root: str, mode: str):
+    img_list = []
+    for img_name in os.listdir(os.path.join(root, mode, 'img')):
+        gt_name = img_name.replace('hazy', 'GT')
+        assert os.path.exist(os.path.join(root, mode, 'gt', gt_name))
+        img_list.append([os.path.join(root, mode, 'img', img_name),
+                         os.path.join(root, mode, 'gt', gt_name)])
+    return img_list
+
+
 def make_dataset_oihaze_train(root, suffix):
     items = []
     for img_name in os.listdir(os.path.join(root, 'haze' + suffix)):
@@ -242,6 +252,34 @@ def __len__(self):
         return len(self.imgs)
 
 
+class OHazeDataset(data.Dataset):
+    def __init__(self, root, mode):
+        self.root = root
+        self.mode = mode
+        self.imgs = make_dataset_ohaze(root, mode)
+
+    def __getitem__(self, index):
+        img_path, gt_path = self.imgs[index]
+        name = os.path.splitext(os.path.split(haze_path)[1])[0]
+
+        img = Image.open(img_path).convert('RGB')
+        gt = Image.open(gt_path).convert('RGB')
+
+        if 'train' in self.mode:
+            # img, gt = random_crop(416, img, gt)
+            if random.random() < 0.5:
+                img = img.transpose(Image.FLIP_LEFT_RIGHT)
+                gt = gt.transpose(Image.FLIP_LEFT_RIGHT)
+
+            rotate_degree = np.random.choice([-90, 0, 90, 180])
+            img, gt = img.rotate(rotate_degree, Image.BILINEAR), gt.rotate(rotate_degree, Image.BILINEAR)
+
+        return to_tensor(img), to_tensor(gt), name
+
+    def __len__(self):
+        return len(self.imgs)
+
+
 class OIHaze(data.Dataset):
     def __init__(self, root, mode, suffix=None, flip=False, crop=None):
         assert mode in ['train', 'test']
diff --git a/train.py b/train.py
index b75a11e..adee1ba 100644
--- a/train.py
+++ b/train.py
@@ -30,6 +30,7 @@ def parse_args():
 
 
 cfgs = {
+    'use_physical': True,
     'iter_num': 40000,
     'train_batch_size': 16,
     'last_iter': 0,