2 ms·
I'm guessing you're not storing the CLIP for every single frame, instead of every second or so? Also, are you using the cosine similarity? How are you finding t
by vhcr 10mo ago
I'm guessing you're not storing the CLIP for every single frame, instead of every second or so? Also, are you using the cosine similarity? How are you finding the nearest vector?
- laidoffamazon 10mo agoI split per scene using pyscenedetect and sampled from each. Distance is via cosine similarity- I fed it into qdrant