refactor(preprocessor): use io utils in audio and gif

gnes/preprocessor/audio/audio_vanilla.py

@@ -16,24 +16,22 @@
 import numpy as np
 
 from ..base import BaseAudioPreprocessor
-from ..helper import get_video_length_from_raw, get_audio
 from ...proto import array2blob
+from ..io_utils.audio import split_audio
 
 
 class AudioVanilla(BaseAudioPreprocessor):
 
-    def __init__(self, audio_interval: int,
+    def __init__(self,
                  sample_rate: int, *args, **kwargs):
         super().__init__(*args, **kwargs)
-        self.audio_interval = audio_interval
         self.sample_rate = sample_rate
 
     def apply(self, doc: 'gnes_pb2.Document') -> None:
         super().apply(doc)
 
         if doc.raw_bytes:
-            audio = get_audio(doc.raw_bytes, self.sample_rate,
-                              self.audio_interval, get_video_length_from_raw(doc.raw_bytes))
+            audio = split_audio(input_data=doc.raw_bytes, sample_rate=self.sample_rate)
             if len(audio) >= 1:
                 for ci, chunks in enumerate(audio):
                     c = doc.chunks.add()

gnes/preprocessor/io_utils/audio.py

@@ -147,6 +147,7 @@ def split_audio(input_fn: str = 'pipe:',
                 silence_duration=DEFAULT_SILENCE_DURATION,
                 start_time: float = None,
                 end_time: float = None,
+                sample_rate: int = 16000,
                 verbose=False):
     _check_input(input_fn, input_data)
     chunk_times = get_chunk_times(
@@ -157,7 +158,7 @@ def split_audio(input_fn: str = 'pipe:',
         start_time=start_time,
         end_time=end_time)
     audio_chunks = list()
-    print("chunk_times", chunk_times)
+    # print("chunk_times", chunk_times)
     for i, (start_time, end_time) in enumerate(chunk_times):
         time = end_time - start_time
         if time < 0:
@@ -169,7 +170,8 @@ def split_audio(input_fn: str = 'pipe:',
         }
 
         output_kwargs = {
-            'format': 'wav'
+            'format': 'wav',
+            'ar': sample_rate
         }
 
         cmd_args = compile_args(

gnes/preprocessor/video/ffmpeg.py

@@ -19,7 +19,7 @@
 import numpy as np
 
 from ..base import BaseVideoPreprocessor, RawChunkPreprocessor
-from ..helper import split_video_frames, phash_descriptor, get_gif
+from ..helper import split_video_frames, phash_descriptor
 from ...proto import gnes_pb2, array2blob, blob2array
 from ..io_utils import video as video_util
 
@@ -187,4 +187,6 @@ def apply(self, doc: 'gnes_pb2.Document') -> None:
 class GifChunkPreprocessor(RawChunkPreprocessor, BaseVideoPreprocessor):
     @staticmethod
     def _parse_chunk(chunk: 'gnes_pb2.Chunk', *args, **kwargs):
-        return get_gif(blob2array(chunk.blob))
+        from ..io_utils import gif as gif_util
+
+        return gif_util.encode_gif(blob2array(chunk.blob), fps=10)