Upload folder using huggingface_hub

.hugignore

@@ -1,6 +1,9 @@
 **/*.zip
+**/*.mpg
+**/*.align
 **/__MACOSX
 GRID-dataset.zip
+GRID-dataset/**
 models/**
 venv/**
 config.yml

.idea/workspace.xml

@@ -5,14 +5,9 @@
   </component>
   <component name="ChangeListManager">
     <list default="true" id="98e215cf-e237-4c1b-a854-08c5c8eeb9db" name="Changes" comment="">
-      <change afterPath="$PROJECT_DIR$/.hugignore" afterDir="false" />
-      <change afterPath="$PROJECT_DIR$/check_videos.py" afterDir="false" />
-      <change afterPath="$PROJECT_DIR$/model.py" afterDir="false" />
-      <change afterPath="$PROJECT_DIR$/upload.py" afterDir="false" />
-      <change beforePath="$PROJECT_DIR$/.gitignore" beforeDir="false" afterPath="$PROJECT_DIR$/.gitignore" afterDir="false" />
-      <change beforePath="$PROJECT_DIR$/Loader.py" beforeDir="false" afterPath="$PROJECT_DIR$/Loader.py" afterDir="false" />
+      <change beforePath="$PROJECT_DIR$/check_videos.py" beforeDir="false" afterPath="$PROJECT_DIR$/check_videos.py" afterDir="false" />
+      <change beforePath="$PROJECT_DIR$/corrupted.txt" beforeDir="false" afterPath="$PROJECT_DIR$/corrupted.txt" afterDir="false" />
       <change beforePath="$PROJECT_DIR$/helpers.py" beforeDir="false" afterPath="$PROJECT_DIR$/helpers.py" afterDir="false" />
-      <change beforePath="$PROJECT_DIR$/train.py" beforeDir="false" afterPath="$PROJECT_DIR$/train.py" afterDir="false" />
     </list>
     <option name="SHOW_DIALOG" value="false" />
     <option name="HIGHLIGHT_CONFLICTS" value="true" />
@@ -72,7 +67,10 @@
       <updated>1698473948341</updated>
       <workItem from="1698473950406" duration="11154000" />
       <workItem from="1698545246834" duration="6111000" />
-      <workItem from="1698579018554" duration="12142000" />
+      <workItem from="1698579018554" duration="14141000" />
+      <workItem from="1698650359117" duration="2013000" />
+      <workItem from="1698671103984" duration="599000" />
+      <workItem from="1698775864388" duration="1213000" />
     </task>
     <servers />
   </component>

check_videos.py

@@ -11,11 +11,11 @@ corrupted_videos = []
 for video_path in tqdm(all_videos):
     try:
         load_video(video_path)
-    except (cv2.error, ValueError) as e:
+    except Exception as e:
         corrupted_videos.append(video_path)
         print('corrupted video', video_path)
 
 print(f'corrupted videos: {len(corrupted_videos)}')
-record_file = open('corrupted.txt', 'w')
+record_file = open('corrupted-videos.txt', 'w')
 record_file.write('\n'.join(corrupted_videos))
 record_file.close()

corrupted.txt

@@ -1,107 +0,0 @@
-GRID-dataset/videos/s8/lgazzs.mpg
-GRID-dataset/videos/s8/lbwx9n.mpg
-GRID-dataset/videos/s8/lwiyzs.mpg
-GRID-dataset/videos/s8/lrwy5n.mpg
-GRID-dataset/videos/s8/pbwv6a.mpg
-GRID-dataset/videos/s8/bgwn6s.mpg
-GRID-dataset/videos/s8/lbip3n.mpg
-GRID-dataset/videos/s8/sgih5n.mpg
-GRID-dataset/videos/s8/lbbx5n.mpg
-GRID-dataset/videos/s8/bgim5p.mpg
-GRID-dataset/videos/s8/brwf9n.mpg
-GRID-dataset/videos/s8/lrak2a.mpg
-GRID-dataset/videos/s8/lbbd6a.mpg
-GRID-dataset/videos/s8/sgbo7n.mpg
-GRID-dataset/videos/s8/pgbj9n.mpg
-GRID-dataset/videos/s8/bwwa2s.mpg
-GRID-dataset/videos/s8/srbh4a.mpg
-GRID-dataset/videos/s8/bganza.mpg
-GRID-dataset/videos/s8/bbie3p.mpg
-GRID-dataset/videos/s8/pwwx7p.mpg
-GRID-dataset/videos/s8/sran1n.mpg
-GRID-dataset/videos/s8/pwwq2s.mpg
-GRID-dataset/videos/s8/pbbb8s.mpg
-GRID-dataset/videos/s8/lbwy2a.mpg
-GRID-dataset/videos/s8/swia6s.mpg
-GRID-dataset/videos/s8/prwc9p.mpg
-GRID-dataset/videos/s8/bgag6a.mpg
-GRID-dataset/videos/s8/prav1n.mpg
-GRID-dataset/videos/s8/braz5p.mpg
-GRID-dataset/videos/s8/sgih8a.mpg
-GRID-dataset/videos/s8/bbir1p.mpg
-GRID-dataset/videos/s8/pgwe1p.mpg
-GRID-dataset/videos/s8/lrbd9n.mpg
-GRID-dataset/videos/s8/pbioza.mpg
-GRID-dataset/videos/s8/lrad8a.mpg
-GRID-dataset/videos/s8/lgbf4a.mpg
-GRID-dataset/videos/s8/sbwn6a.mpg
-GRID-dataset/videos/s8/pwav7n.mpg
-GRID-dataset/videos/s8/lrid4a.mpg
-GRID-dataset/videos/s8/pwii6s.mpg
-GRID-dataset/videos/s8/bwwt5p.mpg
-GRID-dataset/videos/s8/swit9p.mpg
-GRID-dataset/videos/s8/bwaf7n.mpg
-GRID-dataset/videos/s8/sgbv2s.mpg
-GRID-dataset/videos/s8/pbwc1n.mpg
-GRID-dataset/videos/s8/lraq3n.mpg
-GRID-dataset/videos/s8/pgwd9n.mpg
-GRID-dataset/videos/s8/prii2a.mpg
-GRID-dataset/videos/s8/pgbq6a.mpg
-GRID-dataset/videos/s8/pwic2s.mpg
-GRID-dataset/videos/s8/prwx2a.mpg
-GRID-dataset/videos/s8/pwbqza.mpg
-GRID-dataset/videos/s8/brie8s.mpg
-GRID-dataset/videos/s8/sgbi4s.mpg
-GRID-dataset/videos/s8/swwb8s.mpg
-GRID-dataset/videos/s8/prio4s.mpg
-GRID-dataset/videos/s8/sbaaza.mpg
-GRID-dataset/videos/s8/bwbg1n.mpg
-GRID-dataset/videos/s8/pwap4s.mpg
-GRID-dataset/videos/s8/prwp5n.mpg
-GRID-dataset/videos/s8/lbbx6s.mpg
-GRID-dataset/videos/s8/lrid1n.mpg
-GRID-dataset/videos/s8/lbwd8s.mpg
-GRID-dataset/videos/s8/pwic1n.mpg
-GRID-dataset/videos/s8/bbirzs.mpg
-GRID-dataset/videos/s8/swwv2a.mpg
-GRID-dataset/videos/s8/lgar7p.mpg
-GRID-dataset/videos/s8/lbaj5p.mpg
-GRID-dataset/videos/s8/lbbq2s.mpg
-GRID-dataset/videos/s8/lgaz1p.mpg
-GRID-dataset/videos/s8/bgwa8s.mpg
-GRID-dataset/videos/s8/lbip6a.mpg
-GRID-dataset/videos/s8/prao9p.mpg
-GRID-dataset/videos/s8/pwbj4s.mpg
-GRID-dataset/videos/s8/lgir2s.mpg
-GRID-dataset/videos/s8/pwbj6a.mpg
-GRID-dataset/videos/s8/srwb1n.mpg
-GRID-dataset/videos/s8/sgau7n.mpg
-GRID-dataset/videos/s8/bbar6a.mpg
-GRID-dataset/videos/s8/bbay9p.mpg
-GRID-dataset/videos/s8/lwid8s.mpg
-GRID-dataset/videos/s8/pgbd7p.mpg
-GRID-dataset/videos/s8/sgao6a.mpg
-GRID-dataset/videos/s8/bbik5n.mpg
-GRID-dataset/videos/s8/lgiy7p.mpg
-GRID-dataset/videos/s8/lwwl6a.mpg
-GRID-dataset/videos/s8/pwbp8s.mpg
-GRID-dataset/videos/s8/swihzs.mpg
-GRID-dataset/videos/s8/brbf7p.mpg
-GRID-dataset/videos/s8/lwak7p.mpg
-GRID-dataset/videos/s8/brwg2a.mpg
-GRID-dataset/videos/s8/lwwz3p.mpg
-GRID-dataset/videos/s8/prwj3p.mpg
-GRID-dataset/videos/s8/sgai1p.mpg
-GRID-dataset/videos/s8/pbau5n.mpg
-GRID-dataset/videos/s8/swbh8s.mpg
-GRID-dataset/videos/s8/brbz9p.mpg
-GRID-dataset/videos/s8/brbm1p.mpg
-GRID-dataset/videos/s8/srag8s.mpg
-GRID-dataset/videos/s8/lwae1n.mpg
-GRID-dataset/videos/s8/lrip9n.mpg
-GRID-dataset/videos/s8/srbh1n.mpg
-GRID-dataset/videos/s8/lwae4a.mpg
-GRID-dataset/videos/s8/swih2a.mpg
-GRID-dataset/videos/s8/pbab6a.mpg
-GRID-dataset/videos/s8/sbigza.mpg
-GRID-dataset/videos/s8/pwiv5p.mpg

helpers.py

@@ -21,12 +21,23 @@ num_to_char = tf.keras.layers.StringLookup(
 
 
 def load_video(path: str) -> List[float]:
+    """
+    adapted original loading code using this tutorial about openCV
+    https://learnopencv.com/read-write-and-display-a-video-using-opencv-cpp-python/
+    """
     cap = cv2.VideoCapture(path)
     frames = []
-    for _ in range(int(cap.get(cv2.CAP_PROP_FRAME_COUNT))):
+    ret = True
+
+    while cap.isOpened():
         ret, frame = cap.read()
-        frame = tf.image.rgb_to_grayscale(frame)
-        frames.append(frame[190:236, 80:220, :])
+
+        if ret is True:
+            frame = tf.image.rgb_to_grayscale(frame)
+            frames.append(frame[190:236, 80:220, :])
+        else:
+            break
+
     cap.release()
 
     mean = tf.math.reduce_mean(frames)

upload.py

@@ -7,5 +7,5 @@ api.upload_folder(
     folder_path="./",
     repo_id="SilentSpeak/lipnet",
     repo_type="model",
-    ignore_patterns=open('.gitignore').read().split('\n')
+    ignore_patterns=open('.hugignore').read().split('\n')
 )