6 ms·
The one thing I've always wanted from identicons is for the system to use a mixed generator composed of at least 20 different algorithms, that all produce resul
by derefr 7y ago
The one thing I've always wanted from identicons is for the system to use a mixed generator composed of at least 20 different algorithms, that all produce results that are completely different-in-kind (like abstract shapes vs. cartoon monsters vs. anime girls vs. swatched spirographs vs. pixel cities vs. ...), such that it's unlikely that any two anonymous users in a smaller conversation will end up with identicons that you have to inspect to differentiate. (E.g. hopefully, if there are two users with identicons of the same style, then they would have very different base colors, which is impossible if all the identicons are of the same style.)
- visarga 7y agoYou can just train a conditional GAN on a variety of datasets, such as BigGAN. https://medium.com/syncedreview/biggan-a-new-state-of-the-art-in-image-synthesis-cf2ec5694024 https://medium.com/syncedreview/biggan-a-new-state-of-the-ar...