diff --git a/torchvision/datasets/celeba.py b/torchvision/datasets/celeba.py
index 12beded2a19..cc6f8084a80 100644
--- a/torchvision/datasets/celeba.py
+++ b/torchvision/datasets/celeba.py
@@ -40,9 +40,9 @@ class CelebA(VisionDataset):
     # dependencies). The "in-the-wild" (not aligned+cropped) images are only in 7z, so they are not available
     # right now.
     file_list = [
-        # File ID                         MD5 Hash                            Filename
+        # File ID                                      MD5 Hash                            Filename
         ("0B7EVK8r0v71pZjFTYXZWM3FlRnM", "00d2c5bc6d35e252742224ab0c1e8fcb", "img_align_celeba.zip"),
-        # ("0B7EVK8r0v71pbWNEUjJKdDQ3dGc", "b6cd7e93bc7a96c2dc33f819aa3ac651", "img_align_celeba_png.7z"),
+        # ("0B7EVK8r0v71pbWNEUjJKdDQ3dGc","b6cd7e93bc7a96c2dc33f819aa3ac651", "img_align_celeba_png.7z"),
         # ("0B7EVK8r0v71peklHb0pGdDl6R28", "b6cd7e93bc7a96c2dc33f819aa3ac651", "img_celeba.7z"),
         ("0B7EVK8r0v71pblRyaVFSWGxPY0U", "75e246fa4810816ffd6ee81facbd244c", "list_attr_celeba.txt"),
         ("1_ee_0u7vcNLOfNLegJRHmolfH5ICW-XS", "32bd1bd63d3c78cd57e08160ec5ed1e2", "identity_CelebA.txt"),
diff --git a/torchvision/datasets/omniglot.py b/torchvision/datasets/omniglot.py
index 2f20bff72c6..b78bf86d16f 100644
--- a/torchvision/datasets/omniglot.py
+++ b/torchvision/datasets/omniglot.py
@@ -22,7 +22,7 @@ class Omniglot(VisionDataset):
             downloaded again.
     """
     folder = 'omniglot-py'
-    download_url_prefix = 'https://github.com/brendenlake/omniglot/raw/master/python'
+    download_url_prefix = 'https://raw.githubusercontent.com/brendenlake/omniglot/master/python'
     zips_md5 = {
         'images_background': '68d2efa1b9178cc56df9314c21c6e718',
         'images_evaluation': '6b91aef0f799c5bb55b94e3f2daec811'
diff --git a/torchvision/datasets/sbd.py b/torchvision/datasets/sbd.py
index f6d7031a95e..e47c9493858 100644
--- a/torchvision/datasets/sbd.py
+++ b/torchvision/datasets/sbd.py
@@ -41,7 +41,7 @@ class SBDataset(VisionDataset):
             if `mode='boundaries'` or PIL image if `mode='segmentation'`.
     """
 
-    url = "http://www.eecs.berkeley.edu/Research/Projects/CS/vision/grouping/semantic_contours/benchmark.tgz"
+    url = "https://www2.eecs.berkeley.edu/Research/Projects/CS/vision/grouping/semantic_contours/benchmark.tgz"
     md5 = "82b4d87ceb2ed10f6038a1cba92111cb"
     filename = "benchmark.tgz"
 
diff --git a/torchvision/datasets/widerface.py b/torchvision/datasets/widerface.py
index 55ad6d1e76a..c1775309b29 100644
--- a/torchvision/datasets/widerface.py
+++ b/torchvision/datasets/widerface.py
@@ -37,7 +37,7 @@ class WIDERFace(VisionDataset):
 
     BASE_FOLDER = "widerface"
     FILE_LIST = [
-        # File ID                        MD5 Hash                            Filename
+        # File ID                             MD5 Hash                          Filename
         ("0B6eKvaijfFUDQUUwd21EckhUbWs", "3fedf70df600953d25982bcd13d91ba2", "WIDER_train.zip"),
         ("0B6eKvaijfFUDd3dIRmpvSk8tLUk", "dfa7d7e790efa35df3788964cf0bbaea", "WIDER_val.zip"),
         ("0B6eKvaijfFUDbW4tdGpaYjgzZkU", "e5d8f4248ed24c334bbd12f49c29dd40", "WIDER_test.zip")