News for hackers. Stuff that matters.

A comment on ScaNN: Efficient Vector Similarity Search. Parent comment

Re:ScaNN: Efficient Vector Similarity Search by natemwilson on Thursday July 30, 2020 @12:59AM (#23994837)
How would a reverse video search engine work? You give it a clip and it shows you which longer clips it is in?
[ Parent | on HN ]
Re:ScaNN: Efficient Vector Similarity Search by willvarfar on Thursday July 30, 2020 @05:03AM (#23996060)
I would imagine it using a Shazam-like algorithm, only for downsampled video and not just the audio.

A decade ago there was some hobby implementations of these algorithms that got cease and desist letters form Shazam and that was all over HN.

https://m.youtube.com/watch?v=8Dj0rekeM7g

[ Parent | on HN ]
Re:ScaNN: Efficient Vector Similarity Search by Immortal333 on Thursday July 30, 2020 @02:54AM (#23995392)
Yes, Exactly same as you described.
[ Parent | on HN ]
Re:ScaNN: Efficient Vector Similarity Search by jointpdf on Thursday July 30, 2020 @03:43AM (#23995654)
Could you take (embeddings of) frames from your query video and measure the pairwise similarity to frames from the reference videos, then rank by some summary function of that list of similarity scores?

What are the bottlenecks besides performance?

[ Parent | on HN ]