Week 9 2019 Fall

• Reading:
  • tfjs: Handwritten digit recognition with CNNs
• Coding:
  • Play with the doodle classifier examples, try different data, tweak the hyperparameters, modify the architecture of the model.
• Writing:
  • Write down what kind of data, hyperparameters, and architecture you find that work better in this case.

Doodle Classifier Examples

• (1) Doodle Classifier on 100 classes, code: pre-trained model

• (2) Doodle Classifier with KNN Classifier, code: transfer learning

• (3) Train your own Doodle Classifier, code: create our own model

• Doodle Classifier on 345 classes

• Get Quickdraw Dataset

• Train your own MNIST Classifier

Resource:

• Quickdraw data download
• Coding train: Doodle Classifier - Intelligence and Learning
• Convolutional Neural Networks - The Math of Intelligence

Add your name / link to your assignment below

• Your name, [Project name](Project Link), Project Description
• Cara Neel, demo, CNN to determine if you drew an Ear, Elephant or Envelope.