Apr 26, 2020 18:07:00

'dick-rnn' - AI helps you finish drawing a penis

' dick-rnn ' was created as a

recurrent neural network (RNN) model similar to ' Sketch-RNN, ' which uses a neural network to complete an illustration from lines drawn in the middle of the drawing. As its name suggests, 'dick-rnn' is an RNN specialized for illustrations of 'dick.'

[P] I trained a recurrent neural network to draw dick doodles: Machine Learning
https://www.reddit.com/r/MachineLearning/comments/g6og9l/p_i_trained_a_recurrent_neural_network_trained_to/

'dick-rnn' was created by richard, a foreign bulletin board and reddit user. According to him, 'dick-rnn' is a fork of 'Sketch-RNN', which finishes illustrations from lines in the middle of drawing, and the source code is repository on GitHub. The dataset used to train 'dick-rnn' is based on Quickdraw-appendix .

GitHub - dickrnn/dickrnn.github.io: a recurrent network trained to draw dicks
https://github.com/dickrnn/dickrnn.github.io/

In 2018, Google open-sourced the Quick, Draw! dataset. Richard describes this dataset as 'the world's largest doodle dataset.' The Quick, Draw! dataset contains 345 categories and over 50 million illustrations, but Richard points out that 'this dataset does not include specific categories that people enjoy doodles in.' In addition, 'penis prank drawings are a casual symbol of rebellious behavior,' and 'moral guidelines should not be in the hands of major technology companies.' It seems that he decided to develop 'dick-rnn' out of a rebellious spirit that 'moral guidelines should not be in the hands of major technology companies.'

In addition, 'dick-rnn' trains a neural network using 10,000 illustrations of penises, and you can experience its capabilities on the following demo site.

dick-rnn
https://dickrnn.github.io/

The demo site is easy to use; all you have to do is use the mouse to draw lines.

The neural network then draws the rest of the illustration. The lines drawn by the user are displayed in black, while the lines drawn by the neural network are displayed in colorful non-black lines.

If the user does not perform any operations, 'dick-rnn' will continue to draw various penis illustrations that fit the original lines.

In addition, overseas technology media TNW states, 'We realized that the easiest way to help AI draw a real penis is to draw the testicles,' and even when actually using it, it seemed that the illustration was more complete when the testicles were drawn and the continuation was completed than when other parts were drawn and 'dick-rnn' was asked to continue the illustration.