diff --git a/README.md b/README.md
index 607cbbe..65c58a4 100644
--- a/README.md
+++ b/README.md
@@ -18,7 +18,7 @@ The weights and the model are exactly the same as in [the official Tensorflow im
 Install from [pip](https://pypi.org/project/pytorch-fid/):
 
 ```
-pip install pytorch-fid
+pip install git+https://github.com/XavierJiezou/pytorch-fid.git
 ```
 
 Requirements:
diff --git a/src/pytorch_fid/fid_score.py b/src/pytorch_fid/fid_score.py
index 9c8acb2..3736b01 100755
--- a/src/pytorch_fid/fid_score.py
+++ b/src/pytorch_fid/fid_score.py
@@ -33,6 +33,7 @@
 """
 
 import os
+import glob
 import pathlib
 from argparse import ArgumentDefaultsHelpFormatter, ArgumentParser
 
@@ -268,9 +269,7 @@ def compute_statistics_of_path(path, model, batch_size, dims, device, num_worker
             m, s = f["mu"][:], f["sigma"][:]
     else:
         path = pathlib.Path(path)
-        files = sorted(
-            [file for ext in IMAGE_EXTENSIONS for file in path.glob("*.{}".format(ext))]
-        )
+        files = sorted([file for ext in IMAGE_EXTENSIONS for file in glob.glob(os.path.join(path, '**/*.{}'.format(ext)), recursive=True)]) # 支持递归搜索子目录
         m, s = calculate_activation_statistics(
             files, model, batch_size, dims, device, num_workers
         )