Skip to content

Latest commit

 

History

History
30 lines (24 loc) · 890 Bytes

README.md

File metadata and controls

30 lines (24 loc) · 890 Bytes
Raw

Doodle-Classifier

Classify the doodles drawn on the canvas using a Neural Network Library and dataset from the googlecreativelab's quickdraw dataset.
A project inspired from Google Quick Draw game - https://quickdraw.withgoogle.com/.

Dataset

Using googlecreativelab's quickdraw dataset.
https://github.com/googlecreativelab/quickdraw-dataset
MNIST database.
http://yann.lecun.com/exdb/mnist/

Language used.

Processing(Java) - To load and save the data from the quickdraw dataset from numpy(.npy) to binary(.BIN) format. https://processing.rg
p5(JavaScript) - Input Data --> Neural Network Library --> Output . Makes use of the Browser Canvas to read in the doodle. https://p5js.org/

Concepts

  • Supervised Learning
  • Training and Testing Data
  • Overfitting
  • Softmax
  • Cross-entropy