Deliverable 1: Tensorflow implementation of Drawing Classification

This repo contains the TensorFlow code for sketch-rnn, the recurrent neural network model described in Teaching Machines to Draw and A Neural Representation of Sketch Drawings. We train our model on a dataset of hand-drawn sketches, each represented as a sequence of motor actions controlling a pen: which direction to move, when to lift the pen up, and when to stop drawing. You can use the jupyter notebook included to encode, decode, and morph between two vector images, and also generate new random ones.

Requirements

• Python
• Pip package manager
• Tensorflow
• Magenta
• Jupyter Notebook
• NodeJS

Data Set - The Quick, Draw! Dataset

• The Quick Draw Dataset is a collection of 50 million drawings across 345 categories, contributed by players of the game Quick, Draw!. The drawings were captured as timestamped vectors, tagged with metadata including what the player was asked to draw and in which country the player was located.
• The dataset is available on Google Cloud Storage as ndjson files seperated by category.

Instructions

• Download the Magenta implementation of Sketch RNN
• Train the model - sketch_rnn_train --log_root=checkpoint_path --data_dir=dataset_path --hparams="data_set=[dataset_filename.npz]" and save the output checkpoints and weights.
• You can now use your model to view output using Sketch_RNN.ipynb
• Additionally, you use this model for the web by converting into a magenta-js readable format. Sketch RNN TF to JS
• Use json output of previous step in the magenta-js sketch rnn web application. Sketch_RNN_TF_To_JS_Tutorial.ipynb

Source code: Tensorflow implementation of Drawing Classification - ZIP

Github: Tensorflow implementation of Drawing Classification