This project uses a simple ResNet model to identify doodle drawings in online game Skribbl.io
The model used is trained on Google's Quickdraw Dataset
This project uses OpenCV to capture the drawings being drawn by the player, PyTorch as the deep learning framework, and Torchserve to deploy model on localhost.
Some of the outputs of the model being predicted in real-time are shown below:-
Don't focus on the word shown by Skribbl, didn't use custom words for the game
Please don't judge me
There are no steps to reproduce, however if you want to try it out on your own system, feel free to get in touch.