adding PySceneDetect to plier

[yibeichan]

Happy New Year!

This is a follow-up to discussion with @adelavega on Mattermost. In case of shot detection or segment detection, Google API has crushed a lot when dealing with large videos, such as movies in NaturalisticNeuroimagingDatabase. PySceneDetect can be a nice alternative for Google API in this case. Do you want to add PySceneDetect to plier? (I can help to code but probably not until late March 2023

Thank you!

[adelavega]

I will take a look. If it's not too hard I may be able to do it!

[yibeichan]

Sounds good! Btw, do you know Deepface? It integrates mainstream face recognition packages. I first used face_recognition/Dlib to identify characters' faces in movies (NNDb), then found that Deepface intergrates all of them.

[adelavega]

Cool. I did not. In the past I've used facenet (which is based on dlib + other networks), but that looks like a good package.

[yibeichan]

Found this opensource speech recognition package whisper; here is a blog that compares the performance among AssemblyAI, whisper, and Google Speech-to-text. I haven't used this package but I'm happy to give it a try (e.g., transcribe movies in NNDb). (I'm using this issue as a record for these as-potential-substitute packages)

[adelavega]

I've heard great things about whisper!