We present our data preprocessing, models fittings, and a fun application of the Google Quick, Draw! dataset. This project was built by Akhilesh Reddy, Vincent Kuo, Kirti Pande, Tiffany Sung, and Helena Shi. To see the full walk-through, find our blog post at: blog
Walk with us through this journey to see how we have tackled the challenge in successfully classifying what is “arguably the world’s cutest research dataset!”
Shuffle_CSV_and_ResNet folder: includes data pre-processing/shuffle csv code and the SE-ResNet model we tried.
Mobile_net folder: includes the Mobile Net model code and training.
OtherModels folder: we tried some skicit learn classifiers for instance Random Forest, k-NN and MLP Classifier.
QuickDraw folder: One can replicate the application we created by running the QuickDrawApp.py code. Also, download the CNN model from the following link: https://drive.google.com/open?id=1PAFXI5HrY7HguZy0I8yDaUQUebTQYus0 and save in the same location as the folder.
The model is trained for 15 classes:
Donut 🍩
Eye 👁
Tent ⛺
Bicycle 🚲
Flower 🌸
Mermaid 🧜
Snake 🐍
Camera 📷
Pineapple 🍍
Rainbow 🌈
Lipstick 💄
Face 😊
Shoe 👠
Cup ☕
Hockey stick 🏑
A preview of the application is as follows:
