AI Breakout Notes

Interests: - General interest in AI/Applications - Video Filtering (de-noising, sharpening) - A/V fingerprinting - Playback optimizations / bandwidth optimizations / quality optimizations - AV1? encoding optimizations (search paths) - Summarization - AI for ABR

What is the current state of the art for speech to text? - Google, VoiceBase? are state of the art.

Discussion about how to define AI:

    - AI itself has not significantly changed over the years.
    - Datasets have grown, which led to increased popularity.

What is the use case for face detection for IA:

    - Library too big to do manually.
    - Analytical purposes (e.g. did Trump get more air time on tv, can be tracked by detecting his face).

Commonly used tools:

    - OCR: Tesseract,  Abbyy
    - Face Detection: FaceNet? (uses OpenFace?) / OpenCV?
    - Frameworks: TensorFlow?
    - Audio Fingerprinting: audfprint

- We probably have not scratched the surface of what we can do with AI.

- Applying machine learning techniques might be interesting to video encoders.

- Curation of audio descriptions could be powered by scene metadata extracted using machine learning techniques.

- Probably too early for standardization right now:

    - WebVTT Metadata track could do for now.