IO documentation (#106)

* IO documentation

README.md

@@ -83,6 +83,17 @@ Since each time the script is run the list of input files is shuffled, you don't
 workers will be processing the same video.
 On a rare occasion when the collision happens, the script will rewrite previously extracted features.
 
+
+## Input
+The inputs are paths to video files.
+Paths can be passed as a list of paths or as a text file formatted with a single path per line.
+
+
+## Output
+Output is defined by the `on_extraction` argument; by default it prints the features to the command line.
+Possible values of output are ['print', 'save_numpy', 'save_pickle']. `save_*` options save the features in
+the `output_path` folder with the same name as the input video file but with the `.npy` or `.pkl` extension.
+
 ## Used in
 
 * [SpecVQGAN](https://arxiv.org/abs/2110.08791) branch `specvqgan`

utils/utils.py

@@ -131,7 +131,8 @@ def form_list_from_user_input(
         to_shuffle: bool = True,
     ) -> list:
     '''User specifies either list of videos in the cmd or a path to a file with video paths. This function
-       transforms the user input into a list of paths.
+       transforms the user input into a list of paths. Files are expected to be formatted with a single
+       video-path in each line.
 
     Args:
         video_paths (Union[str, ListConfig, None], optional): a list of video paths. Defaults to None.