Quickdraw

GroupID: Group 3

Description:

Quick, Draw! is an online game developed by Google that challenges players to draw a picture of an object or idea and then uses a neural network artificial intelligence to guess what the drawings represent.

Motivation:

To build models that make prediction out of Complete as well as Incomplete sketches and compare the results.

Author

1. Liu Chang
2. Zhang Yu
3. Qiu Yang
4. Ding Shuya

To Run the codes

1. Get raw data: 1102_05b633244.dms https://drive.google.com/drive/folders/1ovQX5gqle7JOL7RcO6f_05n7DW2FrOUk?usp=sharing

2. Run DatasetGeneration.ipynb in the utils folder, to generate data files. use lines in read_data.py in your code to load these files.

3. Pick a model that you want to run, change the data path and you should be ready to go.

All codes and other data are also available at git clone https://github.com/neilding69/quickdraw

Folder Structure

1. Person_Modelname.ipynb：model reproduce
2. Data Analysis: data analysis of quick draw
3. Utils folder: data generation and preprocessing file
4. Toy_dataset folder: some toy dataset (used in qiuyang_lstm_cnn.ipynb)

Model Structure

1. CNN: dingshuya_cnn.ipynb
2. LSTM: zhangyu_lstm.ipynb
3. CNN-GRU:liuchang_CNN_GRU.ipynb
4. LSTM-CNN: dingshuya_lstm_cnn.ipynb
5. LSTM-CNN: zhangyu_lstm_cnn.ipynb
6. LSTM-CNN: qiuyang_lstm_cnn.ipynb
7. CNN-LSTM-CNN: qiuyang_cnn_lstm_cnn.ipynb

Results: