You can add location information to your Tweets, such as your city or precise location, from the web and via third-party applications. You always have the option to delete your Tweet location history. Learn more
Purple is close in spectral distance to many common colors but not a common color itself? I wonder if adding a term that penalizes based on prior probability of a color might help. Or does the model already have something like that?
There isn't a loss term like that currently but I'm really reluctant to add anything like that just because it seems to me like that could really be difficult to do properly (unintended negative effects basically)...
Understandable. It could be a lot of work... maybe next version or next project instead :)
I have a suggestion that might help with the purple colors. Break your unet model into 2 parts. The first model encodes the gist of the image, then apply the output to the layers of the unet like stylegan. Use a small multiplier to update the gradients of the 1st net less often.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.
