VideoLearner inside utils_cv.action_recognition.model [ASK]

[usworked]

Description

I'm looking at the Action Recognition model by loading a video file. It seems VideoLearner is using predict_video. Is there a way to use this to split or trim a video based off the classes? In other words, exporting/saving multiple video files from one based off the class presented?

[PatrickBue]

Not 100% what you mean. In its current version we do not support trimming of videos (e.g. ffmpeg could maybe be used for that)

[usworked]

Sorry, let me rephrase.

Currently, using Jupyter, my last line of code is

learner.predict_video(
    video,
    LABELS,
    averaging_size=AVERAGING_SIZE,
    score_threshold=SCORE_THRESHOLD,
    target_labels=TARGET_LABELS,
)

How do I loop over all the prediction labels so I can cut out subclips?

[PatrickBue]

Fyi @jiata