diff --git a/codes/data/base_unsupervised_image_dataset.py b/codes/data/base_unsupervised_image_dataset.py
index 8391168e..20f2805b 100644
--- a/codes/data/base_unsupervised_image_dataset.py
+++ b/codes/data/base_unsupervised_image_dataset.py
@@ -17,6 +17,7 @@ class BaseUnsupervisedImageDataset(data.Dataset):
         self.for_eval = opt['eval'] if 'eval' in opt.keys() else False
         self.scale = opt['scale'] if not self.for_eval else 1
         self.paths = opt['paths']
+        self.corrupt_before_downsize = opt['corrupt_before_downsize'] if 'corrupt_before_downsize' in opt.keys() else False
         assert (self.target_hq_size // self.scale) % self.multiple == 0  # If we dont throw here, we get some really obscure errors.
         if not isinstance(self.paths, list):
             self.paths = [self.paths]
@@ -98,6 +99,8 @@ class BaseUnsupervisedImageDataset(data.Dataset):
     def synthesize_lq(self, hs, hrefs, hmasks, hcenters):
         h, w, _ = hs[0].shape
         ls, lrs, lms, lcs = [], [], [], []
+        if self.corrupt_before_downsize and not self.for_eval:
+            hs = self.corruptor.corrupt_images(hs)
         for hq, hq_ref, hq_mask, hq_center in zip(hs, hrefs, hmasks, hcenters):
             if self.for_eval:
                 ls.append(hq)
@@ -110,7 +113,7 @@ class BaseUnsupervisedImageDataset(data.Dataset):
                 lms.append(cv2.resize(hq_mask, (h // self.scale, w // self.scale), interpolation=cv2.INTER_AREA))
                 lcs.append(self.resize_point(hq_center, (h, w), ls[0].shape[:2]))
         # Corrupt the LQ image (only in eval mode)
-        if not self.for_eval:
+        if not self.corrupt_before_downsize and not self.for_eval:
             ls = self.corruptor.corrupt_images(ls)
         return ls, lrs, lms, lcs
 
diff --git a/codes/data/single_image_dataset.py b/codes/data/single_image_dataset.py
index 9fa2b038..2521f792 100644
--- a/codes/data/single_image_dataset.py
+++ b/codes/data/single_image_dataset.py
@@ -50,9 +50,10 @@ if __name__ == '__main__':
         'force_multiple': 32,
         'scale': 2,
         'eval': False,
-        'fixed_corruptions': ['jpeg-broad'],
+        'fixed_corruptions': ['jpeg-broad', 'gaussian_blur'],
         'random_corruptions': ['noise-5', 'none'],
-        'num_corrupts_per_image': 1
+        'num_corrupts_per_image': 1,
+        'corrupt_before_downsize': True,
     }
 
     ds = SingleImageDataset(opt)
@@ -61,7 +62,7 @@ if __name__ == '__main__':
     for i in range(0, len(ds)):
         o = ds[random.randint(0, len(ds))]
         #for k, v in o.items():
-        k = 'GT'
+        k = 'LQ'
         v = o[k]
         #if 'LQ' in k and 'path' not in k and 'center' not in k:
         #if 'full' in k:
diff --git a/codes/scripts/extract_subimages_with_ref.py b/codes/scripts/extract_subimages_with_ref.py
index 0e058369..8c0460c1 100644
--- a/codes/scripts/extract_subimages_with_ref.py
+++ b/codes/scripts/extract_subimages_with_ref.py
@@ -13,20 +13,20 @@ import torch
 def main():
     split_img = False
     opt = {}
-    opt['n_thread'] = 2
+    opt['n_thread'] = 8
     opt['compression_level'] = 90  # JPEG compression quality rating.
     # CV_IMWRITE_PNG_COMPRESSION from 0 to 9. A higher value means a smaller size and longer
     # compression time. If read raw images during training, use 0 for faster IO speed.
 
     opt['dest'] = 'file'
-    opt['input_folder'] = 'F:\\4k6k\\datasets\\ns_images\\vr\\images_sized'
-    opt['save_folder'] = 'F:\\4k6k\\datasets\\ns_images\\vr\\paired_images'
+    opt['input_folder'] = 'F:\\4k6k\\datasets\\ns_images\\imagesets\\imgset3'
+    opt['save_folder'] = 'F:\\4k6k\\datasets\\ns_images\\imagesets\\256_with_ref_3'
     opt['crop_sz'] = [512, 1024]  # the size of each sub-image
     opt['step'] = [512, 1024]  # step of the sliding crop window
     opt['thres_sz'] = 128  # size threshold
     opt['resize_final_img'] = [.5, .25]
     opt['only_resize'] = False
-    opt['vertical_split'] = True
+    opt['vertical_split'] = False
 
     save_folder = opt['save_folder']
     if not osp.exists(save_folder):
@@ -178,10 +178,14 @@ class TiledDataset(data.Dataset):
     def get(self, index, split_mode, left_img):
         path = self.images[index]
         img = cv2.imread(path, cv2.IMREAD_UNCHANGED)
+
+        if img is None or len(img.shape) == 2:
+            return None
+
         h, w, c = img.shape
 
         # Uncomment to filter any image that doesnt meet a threshold size.
-        if min(h,w) < 1024:
+        if min(h,w) < 512:
             return None
         # Greyscale not supported.
         if len(img.shape) == 2:
diff --git a/codes/scripts/srflow_latent_space_playground.py b/codes/scripts/srflow_latent_space_playground.py
new file mode 100644
index 00000000..e69de29b