diff --git a/PaddleCV/PaddleVideo/data/dataset/kinetics/video2pkl.py b/PaddleCV/PaddleVideo/data/dataset/kinetics/video2pkl.py
index 0950b01b99d156e73b700526fcf123b89300544b..a97ae05358399bc739e3f2476369b695c484988d 100644
--- a/PaddleCV/PaddleVideo/data/dataset/kinetics/video2pkl.py
+++ b/PaddleCV/PaddleVideo/data/dataset/kinetics/video2pkl.py
@@ -68,7 +68,7 @@ def generate_pkl(entry):
     images = sorted(glob.glob(vid + '/*.jpg'))
     ims = []
     for img in images:
-        f = open(img)
+        f = open(img, 'rb')
         ims.append(f.read())
         f.close()
 
diff --git a/PaddleCV/PaddleVideo/data/dataset/youtube8m/tf2pkl.py b/PaddleCV/PaddleVideo/data/dataset/youtube8m/tf2pkl.py
index 7184850b06d57eaac660b1251426761606e3496e..5a559da18a48010588775d48fe07303286ce4c2f 100644
--- a/PaddleCV/PaddleVideo/data/dataset/youtube8m/tf2pkl.py
+++ b/PaddleCV/PaddleVideo/data/dataset/youtube8m/tf2pkl.py
@@ -263,7 +263,7 @@ def main(files_pattern):
             coord.request_stop()
             coord.join(threads)
 
-    print vid_num
+    print(vid_num)
 
     record_name = files_pattern.split('/')[-1].split('.')[0]
     outputdir = target_dir