feature/filter-out-invalid-captions

cdp_backend/pipeline/event_gather_pipeline.py

@@ -160,7 +160,11 @@ def create_event_gather_flow(
                 if session.caption_uri is not None:
                     # If the caption doesn't exist, remove the property
                     # This will result in Speech-to-Text being used instead
-                    if not resource_exists(session.caption_uri):
+                    if not resource_exists(
+                        session.caption_uri
+                    ) or not file_utils.caption_is_valid(
+                        tmp_video_filepath, session.caption_uri
+                    ):
                         log.warning(
                             f"File not found using provided caption URI: "
                             f"'{session.caption_uri}'. "

cdp_backend/tests/utils/test_file_utils.py

@@ -5,9 +5,11 @@
 import random
 import sys
 from pathlib import Path
+from tempfile import TemporaryDirectory
 from typing import Optional
 from unittest import mock
 
+import ffmpeg
 import imageio
 import pytest
 from py._path.local import LocalPath
@@ -16,6 +18,7 @@
 from cdp_backend.utils.file_utils import (
     MAX_THUMBNAIL_HEIGHT,
     MAX_THUMBNAIL_WIDTH,
+    caption_is_valid,
     resource_copy,
 )
 
@@ -312,3 +315,43 @@ def test_clip_and_reformat_video(
     assert outfile.exists()
     assert outfile == expected_outfile
     os.remove(outfile)
+
+
+@pytest.mark.parametrize(
+    "video_uri, caption_uri, end_time, is_resource, expected",
+    [
+        # the video is about 3 minutes and boston_captions.vtt is about 1 minute
+        (EXAMPLE_VIDEO_FILENAME, "boston_captions.vtt", 120, True, False),
+        (EXAMPLE_VIDEO_FILENAME, "boston_captions.vtt", 60, True, True),
+        (
+            EXAMPLE_VIDEO_FILENAME,
+            # about 30 seconds
+            "https://gist.github.com/dphoria/d3f35b5509b784ccd14b7efdc67df752/raw/"
+            "c18fc459c62ff7530536ba19d08021682627c18a/sample.vtt",
+            30,
+            False,
+            True,
+        ),
+    ],
+)
+def test_caption_is_valid(
+    resources_dir: Path,
+    video_uri: str,
+    caption_uri: str,
+    end_time: int,
+    is_resource: bool,
+    expected: bool,
+) -> None:
+    def path_as_str(path: Path) -> str:
+        return str(bytes(path), encoding="utf-8")
+
+    with TemporaryDirectory() as dir_path:
+        temp_video = path_as_str(Path(dir_path) / f"caption-test-{end_time}.mp4")
+        ffmpeg.input(path_as_str(resources_dir / video_uri)).output(
+            temp_video, codec="copy", t=end_time
+        ).run(overwrite_output=True)
+
+        if is_resource:
+            caption_uri = path_as_str(resources_dir / caption_uri)
+
+        assert caption_is_valid(temp_video, caption_uri) == expected

cdp_backend/utils/file_utils.py

@@ -7,13 +7,21 @@
 import re
 from hashlib import sha256
 from pathlib import Path
+from tempfile import TemporaryDirectory
 from typing import Optional, Tuple, Union
 from uuid import uuid4
 
+import ffmpeg
 import fireo
 import fsspec
 import requests
+import webvtt
 from fsspec.core import url_to_fs
+from webvtt.exceptions import (
+    InvalidCaptionsError,
+    MalformedCaptionError,
+    MalformedFileError,
+)
 
 from ..database import models as db_models
 
@@ -626,3 +634,50 @@ def clip_and_reformat_video(
         log.error(ffmpeg_stderr)
 
     return output_path
+
+
+def caption_is_valid(video_uri: str, caption_uri: str) -> bool:
+    """
+    Validate the caption file at the URI provided.
+
+    Parameters
+    ----------
+    video_uri: str
+        The URI for the the target video.
+    uri: str
+        The URI to validate caption file for.
+
+    Returns
+    -------
+    status: bool
+        The validation status.
+
+    Notes
+    -----
+    Duration of the video at video_uri
+    and the duration of the caption file are compared.
+    The caption file is accepted if the durations differ by no more than 20%.
+    """
+    try:
+        ffprobe = ffmpeg.probe(video_uri)
+    except ffmpeg.Error as e:
+        log.warning(f"ffprobe({video_uri}): {e.stderr}")
+        return False
+
+    # Making sure temp copy of the caption file is deleted when finished
+    with TemporaryDirectory() as dir_path:
+        local_caption_path = resource_copy(caption_uri, dst=dir_path)
+        try:
+            caption_length = webvtt.read(local_caption_path).total_length
+        except (InvalidCaptionsError, MalformedCaptionError, MalformedFileError) as e:
+            log.warning(f"webvtt.read({caption_uri}): {str(e)}")
+            return False
+
+        similar_audio_streams = filter(
+            lambda s: s.get("codec_type", "") == "audio"
+            and math.isclose(
+                float(s.get("duration", "0.0")), caption_length, rel_tol=0.2
+            ),
+            ffprobe.get("streams", []),
+        )
+        return any(similar_audio_streams)