Closed Domingowen closed 4 years ago
Hello, This repo is related to the Art-Palette experiment from Google Arts & Culture. The Art Selfie App is using different technics and is not related to Art Palette.
Thanks for your interest!
Hi - I have been using this to match images to a target palette. What should the distance be for an image with itself? I get 100, yet i do find images that have closer distances? Cheers!
Hello, The embedding is using machine learning to speed up the distance calculation. If you want the exact distance you can calculate the exact LAB colors https://en.wikipedia.org/wiki/CIELAB_color_space distance : first order the colors in the palette vectors and then calculate the distance between each ordered vectors. If you have lot of elements to compare and want a fast response you can use the python code to calculate the distance https://github.com/googleartsculture/art-palette/blob/master/palette-embedding/python/palette_embedding.py#L257 and get the nearest neighbors. Simon
Le lun. 6 juil. 2020 à 18:47, zamia-git notifications@github.com a écrit :
Hi - I have been using this to match images to a target palette. What should the distance be for an image with itself? I get 100, yet i do find images that have closer distances? Cheers!
— You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub https://github.com/googleartsculture/art-palette/issues/1#issuecomment-654349298, or unsubscribe https://github.com/notifications/unsubscribe-auth/AACJXADQ235DIYLEA54KPV3R2H53ZANCNFSM4JSJWMFQ .
-- Simon Doury http://www.simondoury.com
ah - perfect - cheers!
I want to use google art historic selfie API match who looks like you in china, I don't have any commercial or never earn money by this action, Just want to this technology to share china, Think about people will like it.