Skip to content

Classifying Google Quick, Draw! Dataset Using Tensoflow 2.0

Notifications You must be signed in to change notification settings

mutexlocker/quickdrawTF2

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

19 Commits
 
 
 
 
 
 

Repository files navigation

Quick, Draw!

This repository contains code for training and inference different neural networks on Google's Quick, Draw! Dataset.

Dataset

Download numpy bitmap files to the data directory. These files contain 28*28 image data of doodle drawings.

cd logs/
gsutil -m cp gs://quickdraw_dataset/full/numpy_bitmap/*.npy .

Networks

Implementation has multiple networks. The best results were achieved by cnn_model3 with top5 category accuracy of 95%.

Config

By default, the code trains for 100 classes and loads 15000 samples for each class. You can change the number of classes to train by changing the value of totalclasses and number of samples for each class by changing the value of samples

Execution

By default the code runs cnn_model3 but can be changed by this line of code:

model_cnn,tensorboard = cnn_model3()

chnage cnn_model3() to any other networks in the list:

  • cnn_model_leaky() CNN with leaky RELU activation
  • cnn_model_lstm() LSTM Network

Logging

Tensoboard is also supported. logs are writtein in logs folder. Run tensoboard as follows:

tensorboard --logdir logs

Save Network

The trained neural network will be saved as two separate file (model.h5 and model.json) when training is completed.

About

Classifying Google Quick, Draw! Dataset Using Tensoflow 2.0

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages