Inspiration
One day we came across a very amazing platform called Google Quick draw and We got this Idea of making Doodle recognition project from there .
What it does
we have built a program that will act as a multi-class classifier to assign hand drawn doodles to a certain domain
How we built it
We built it using Python language , using TensorFlow and Keras Modules
Challenges we ran into
The initial challenges we faced was learning about the implementation of a CNN and what sort of different layers we should create the CNN with to achieve the best accuracy
Accomplishments that we're proud of
Our doodle classifier is working with an acurracy of around 90%
What we learned
We learned about the CNN and its different layers and have also gained more experience in working with modules like opencv2, keras and Tensorflow
What's next for Doodle Recognition
For future work we are planning to experiment our model with more advanced neural networks such as ResNet and also we would prefer using a bigger dataset in future as what we have used today. as we believe that training on bigger dataset will improve the machine's accuracy a lot
Log in or sign up for Devpost to join the conversation.