hardmaru
hardmaru
Cool! I think my implementation is not exactly correct as I had some difficulties understanding the paper. If you want to discuss more later on, feel free to swing by...
Thanks.
Hi @ShusenTang Thanks for the suggestion. When TF was in the 1.2 days the casting was limited so I used 0/1 but your suggestion works well for current TF versions...
Hi @yemeijuan Sketch-RNN doesn't work on raster images. It is a generative model for vector images.
Hi Recently I’m also in the process of porting sketchrnn to TensorFlow.js using TypeScript which is more or less a rewrite with webgl acceleration. The newer version also uses the...
That's a good point, and thanks for the catch. And in fact, in the Sketch-RNN paper, we have changed the definition to use sigma * sqrt(\tau), but have not updated...
In the footnote "Temperature is most commonly discussed in Boltzmann distributions, but can be generalized to all probability distributions", it is not implied that the temperature definition used here *is*...
I think so Although depending on the complexity of the task maybe conventional fixed rate neuroevolution with a fixed network, or reinforcement learning may work more robustly for many tasks...
Hi! Not sure if this helps, but I wrote some scripts in this [repo](https://github.com/hardmaru/sketch-rnn-datasets) that you can use to convert the .npz vector sketches to .svg files, and also display...
It is flip flops. On Fri, Jan 11, 2019 at 11:55 AM maximumnz1 wrote: > It is not flip flops, it was invented by NZ and we named them jandals,...