Page 274 - Artificial Intellegence_v2.0_Class_9
P. 274
Write something using mouse to understand the working of the neural networks.
When you start writing, the experiment suggests you the next letter of the word that you want to write.
4. Variational Auto Encoder
This experiment is based on recurrent neural network (RNN). It creates strokes based on the object you
drawn. It is trained on thousands of human-drawn images.
To play this experiment, visit the following link:
https://magenta.tensorflow.org/assets/sketch_rnn_demo/multi_vae.html
Select a category and draw a shape related to the category in the left-hand side blank area and then
click on the auto-encode button. The related drawing of the object that you have drawn are visible in the
right-hand side cells.
Select different categories and draw an object that is not related to the selected categories and then
see the different outputs.
272 Touchpad Artificial Intelligence (Ver. 2.0)-IX

