feat: add two exemplar scripts

Fairly basic, could obviously be refactored but they are there to be exemplars
to give folks an idea of where they can start with the pipeline.

make_highlight_clips.py

@@ -0,0 +1,97 @@
+"""make_highlight_clips.py -- use the pipeline to make highlight clips from a video
+
+Usage:
+    make_highlight_clips.py <input_file> [--output-dir=<dir>] [--output-filename=<filename>] [--save-json]
+
+Options:
+    -h --help                   Show this screen
+    --output-dir=<dir>          Directory to save the highlight clips [default: /tmp/]
+    --save-json                 Save the feature data as a JSON file [default: False]
+
+Notes:
+    <input_file>                The input file can be a video or JSON file
+"""
+
+import os
+import sys
+import time
+from pipeline.feature_extractors import RandomFeatureExtractor,\
+    LaughterFeatureExtractor, LoudAudioFeatureExtractor,\
+    VideoActivityFeatureExtractor, JSONFeatureExtractor
+from pipeline.utils import SourceMedia, Source
+from pipeline.producers import FfmpegVideoProducer, JSONProducer
+from pipeline.consolidators import OverlapConsolidator
+
+from docopt import docopt
+
+def main():
+    start_time = time.time()
+    PRINT_SOMETHING = False
+    if PRINT_SOMETHING:
+        print(f"+{'-'*78}+")
+        print(f"| Starting pipeline to make highlight clips at {time.strftime('%Y-%m-%d %H:%M:%S')}{' '*13}|")
+        print(f"+{'-'*78}+")
+        time.sleep(1)
+
+    args = docopt(__doc__)
+    input_file = args['<input_file>']
+    output_dir = args['--output-dir']
+    # expand "./" to current directory
+    if output_dir == "./":
+        output_dir = os.getcwd()
+    save_json = args['--save-json']
+
+    source_media = SourceMedia(sources=[Source(input_file, input_file, "manual")])
+
+    combined_features = []
+
+    # feature extractors: laughter, loudness, video activity, words
+    if input_file.endswith(".json"):
+        feature_extractors = [JSONFeatureExtractor]
+    else:
+        feature_extractors = [LaughterFeatureExtractor, LoudAudioFeatureExtractor,
+                              VideoActivityFeatureExtractor
+                              ]
+    # NOTE: words FE crashes OOM due to noted issues on their GH
+    # so omitted for now
+
+    for fe in feature_extractors:
+        fe_instance = fe(input_files=source_media)
+        fe_instance.setup()
+        fe_instance.run()
+        fe_instance.teardown()
+        if fe_instance.features:
+            combined_features.extend(fe_instance.features)
+
+    if combined_features:
+        if PRINT_SOMETHING:
+            for f in combined_features:
+                print(f"Feature: {f}")
+
+        # consolidate features
+        oc = OverlapConsolidator(features=combined_features)
+        oc.consolidate()
+
+        # output before video
+        if save_json:
+            jsonprod = JSONProducer(features=combined_features)
+            jsonprod.produce()
+
+        # make video from features:
+        ffprod = FfmpegVideoProducer(features=oc.features,
+                                     output_dir=output_dir,
+                                     compile_clips=False)
+        ffprod.produce()
+
+
+
+    end_time = time.time()
+    if PRINT_SOMETHING:
+        print(f"+{'-'*78}+")
+        print(f"| Pipeline finished creating highlight video at {time.strftime('%Y-%m-%d %H:%M:%S')}{' '*12}|")
+        print(f"| Total time: {end_time - start_time:.2f} seconds {' '*51}|")
+        print(f"+{'-'*78}+")
+
+
+if __name__ == '__main__':
+    main()

make_highlight_video.py

@@ -0,0 +1,106 @@
+"""make_highlight_video.py -- use the pipeline to make a highlight from a video
+
+Usage:
+    make_highlight_clip.py <input_file> [--max-duration=<seconds>] [--output-dir=<dir>] [--output-filename=<filename>] [--save-json]
+
+Options:
+    -h --help                   Show this screen
+    --max-duration=<seconds>    Maximum duration of the highlight clip [default: 60]
+    --output-dir=<dir>          Directory to save the highlight clip [default: ./]
+    --output-filename=<filename>    Filename of the highlight clip [default: highlights.mp4]
+    --save-json                 Save the feature data as a JSON file [default: False]
+
+Notes:
+    <input_file>                The input file can be a video or JSON file
+"""
+
+import os
+import sys
+import time
+from pipeline.feature_extractors import RandomFeatureExtractor,\
+    LaughterFeatureExtractor, LoudAudioFeatureExtractor,\
+    VideoActivityFeatureExtractor, JSONFeatureExtractor
+from pipeline.utils import SourceMedia, Source
+from pipeline.producers import FfmpegVideoProducer, JSONProducer
+from pipeline.consolidators import OverlapConsolidator
+from pipeline.adjusters import TargetTimeAdjuster
+
+from docopt import docopt
+
+def main():
+    start_time = time.time()
+    PRINT_SOMETHING = False
+    if PRINT_SOMETHING:
+        print(f"+{'-'*78}+")
+        print(f"| Starting pipeline to make highlight video at {time.strftime('%Y-%m-%d %H:%M:%S')}{' '*13}|")
+        print(f"+{'-'*78}+")
+        time.sleep(1)
+
+    args = docopt(__doc__)
+    input_file = args['<input_file>']
+    max_duration = float(args['--max-duration'])
+    output_dir = args['--output-dir']
+    # expand "./" to current directory
+    if output_dir == "./":
+        output_dir = os.getcwd()
+    output_filename = args['--output-filename']
+    save_json = args['--save-json']
+
+    source_media = SourceMedia(sources=[Source(input_file, input_file, "manual")])
+
+    combined_features = []
+
+    # feature extractors: laughter, loudness, video activity, words
+    if input_file.endswith(".json"):
+        feature_extractors = [JSONFeatureExtractor]
+    else:
+        feature_extractors = [LaughterFeatureExtractor, LoudAudioFeatureExtractor,
+                              VideoActivityFeatureExtractor
+                              ]
+    # NOTE: words FE crashes OOM due to noted issues on their GH
+    # so omitted for now
+
+    for fe in feature_extractors:
+        fe_instance = fe(input_files=source_media)
+        fe_instance.setup()
+        fe_instance.run()
+        fe_instance.teardown()
+        if fe_instance.features:
+            combined_features.extend(fe_instance.features)
+
+    if combined_features:
+        if PRINT_SOMETHING:
+            for f in combined_features:
+                print(f"Feature: {f}")
+
+        # consolidate features
+        oc = OverlapConsolidator(features=combined_features)
+        oc.consolidate()
+
+        # adjust features to target time
+        tta = TargetTimeAdjuster(features=oc.features, target_time=max_duration)
+        tta.adjust()
+
+        # output before video
+        if save_json:
+            jsonprod = JSONProducer(features=combined_features)
+            jsonprod.produce()
+
+        # make video from tta features:
+        ffprod = FfmpegVideoProducer(features=tta.features,
+                                     output_dir=output_dir,
+                                     output_filename=output_filename)
+        ffprod.produce()
+
+
+
+    end_time = time.time()
+    if PRINT_SOMETHING:
+        print(f"+{'-'*78}+")
+        print(f"| Pipeline finished creating highlight video at {time.strftime('%Y-%m-%d %H:%M:%S')}{' '*12}|")
+        print(f"| Total time: {end_time - start_time:.2f} seconds {' '*51}|")
+        print(f"+{'-'*78}+")
+
+
+if __name__ == '__main__':
+    main()