Wrapping up

Image Modeling with Keras

Ariel Rokem

Senior Data Scientist, University of Washington

What did we learn?

• Image classification
• Convolutions
• Reducing the number of parameters
  • Tweaking your convolutions
  • Adding pooling layers
• Improving your network
  • Regularization
• Understanding your network
  • Monitoring learning
  • Interpreting the parameters

Model interpretation

https://distill.pub/2017/feature-visualization/

What next?

• Even deeper networks
• Residual networks

Residual networks

What next?

• Even deeper networks
• Residual networks
• Transfer learning
• Fully convolutional networks

Fully convolutional networks

Generative adversarial networks

What next?

• Even deeper networks
• Residual networks
• Transfer learning
• Fully convolutional networks
• Generative adversarial networks
• ...

Good luck!

Image Modeling with Keras