[common] (twitch-dset) Add scripts dir, and stop using watchdog to monitor files that appear

common/research/dataset/twitch/make_thumbnails.py

@@ -48,10 +48,3 @@ class ThumbnailsGenerator:
 
     def _get_frame_number(self):
         return int(ffmpeg.probe(str(self.video_path))['format']['duration'].split('.')[0])
-
-
-if __name__ == '__main__':
-    _video_name = sys.argv[1]
-    print(f'Fragmenting video {_video_name}')
-    ThumbnailsGenerator(_video_name).run()
-

common/research/dataset/twitch/is_robot_view.py → common/research/dataset/twitch/robot_view.py

@@ -5,13 +5,14 @@ from shutil import move
 import numpy as np
 from scipy.spatial import distance
 from skimage import io
-from watchdog.events import FileSystemEvent, FileSystemEventHandler
-from watchdog.observers import Observer
+from skimage.transform import resize
 
 from research.constants import TWITCH_DSET
 
+RES_DIR: Path = TWITCH_DSET / 'robots-views'
+RES_DIR.mkdir(parents=True, exist_ok=True)
 
-ref_image = io.imread('mask.jpg')
+ref_image = io.imread(f'{__file__}/../mask.jpg')
 
 _MASK = ref_image[:, :, 1] > 50
 _REF_IMG_MASKED = ref_image*_MASK[:, :, np.newaxis]
@@ -24,30 +25,14 @@ def is_image_from_robot_view(path_to_image: Path) -> bool:
     return distance.euclidean(img_masked.flatten() / 255, _REF_IMG_MASKED.flatten() / 255) < _THRESHOLD
 
 
-class NewFrameHandler(FileSystemEventHandler):
+def process_image(path_to_image: Path):
+    if is_image_from_robot_view(path_to_image):
+        res_path = RES_DIR / f'{path_to_image.parent.name}-{path_to_image.name}'
+        move(str(path_to_image), str(res_path))
+    else:
+        remove(str(path_to_image))
 
-    def __init__(self):
-        self.res_dir = (TWITCH_DSET / 'robots-views')
-        self.res_dir.mkdir(exist_ok=True, parents=True)
 
-    def on_created(self, event: FileSystemEvent):
-        if event.is_directory or not event.src_path.endswith('jpg'):
-            return
-        file_path = Path(event.src_path)
-        if is_image_from_robot_view(file_path):
-            res = self.res_dir / f'{file_path.parent.name}-{file_path.name}'
-            move(file_path, res)
-        else:
-            remove(file_path)
-
-    def __hash__(self):
-        return hash(self.__class__.__name__)
-
-
-if __name__ == '__main__':
-    obs = Observer()
-    obs.schedule(NewFrameHandler(), str(TWITCH_DSET / 'raw-frames'), recursive=True)
-    obs.start()
-
-    while 1:
-        pass
+def process_all_images_in_dir(dir_path: Path):
+    for path_to_image in dir_path.glob('*/*.jpg'):
+        process_image(path_to_image)

common/research/scripts/monitor_new_twitch_frames.py

+from time import sleep
+
+from research.constants import TWITCH_DSET
+from research.dataset.twitch.robot_view import process_all_images_in_dir
+
+if __name__ == '__main__':
+    while 1:
+        process_all_images_in_dir(TWITCH_DSET / 'raw-frames')
+        sleep(.1)

common/research/scripts/split_video.py

+import sys
+
+from research.dataset.twitch.make_thumbnails import ThumbnailsGenerator
+
+if __name__ == '__main__':
+    _video_name = sys.argv[1]
+    print(f'Fragmenting video {_video_name}')
+    ThumbnailsGenerator(_video_name).run()