Deep Learning

fast.ai

Making neural nets uncool again

Make Deep Learning More accesible

Top down learning

AI,Machine Learning,Deep Learning

Agenda

Cats vs Dogs

Deep Learning Methodology



Training files

Testing Files

How do we do validation

Take sample of training files and use them for validation

Work with sample data first

Summary

  • Divide data into training, validation and testing sets
  • Work with sample of data

Fine Tuning

Fine tuning is a process to take a network model that has already been trained for a given task, and make it perform a second similar task.

VGG16

Model trained on imagenet.

Won ImageNet Large Scale Visual Recognition Competition in 2014.

Live Demo

Sample notebook

To summarize

  • Deep learning Using Keras
  • Deep learning methodologies
  • Finetuning
  • Vgg16
  • Imagenet