search

ClarityCafe / Aya

your globally distributed waifu storage
https://aya.claritycafe.now.sh
MIT License
38 stars 9 forks source link

Smart Duplication Detection #8

Open sr229 opened 4 years ago

sr229 commented 4 years ago

We can use a AI Service for this.

We train a model everytime a new image is uploaded, but its only scheduled per 2 hours to train from the entire collection or just what got uploaded new.

DonaldTsang commented 4 years ago

Also what kind of hashing will you use? There are many like pHash, aHash, dHash and wHash (with varying resolutions)

sr229 commented 4 years ago

I don't think Hashing can solve duplication since anyone can just waifu2x or change the format and it'll result in a different hashing.

DonaldTsang commented 4 years ago

No, pHash and the like hashes the LOOKS of an image, not the data. https://en.wikipedia.org/wiki/Perceptual_hashing

sr229 commented 4 years ago

I'll see what I can do to adapt the hashing standard, but this is on hold for now since its in the FVP log.