Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Deep Learning: Building An Image Classification Model

Deep Learning: Building An Image Classification Model

Grasp insights in Deep Learning by building an image classification network with Convolutional Neural Network

Jennifer Kwentoh

July 21, 2018
Tweet

More Decks by Jennifer Kwentoh

Other Decks in Technology

Transcript

  1. Before we get to the Onion Machine learning Neural Networks

    Deep Learning Convolutional Neural Network Building an Image Classification Model
  2. Machine learning is not bunch of IF Statements To teach

    a machine to recognize a face, we would have to hard code functions of every detail of a face Whereas in machine learning we define the outcome (the face) and the program learns how to get to the output Standard programming Machine Learning
  3. Let’s take a step back to Machine Learning Supervised Learning:

    we give the machine a structured label (outcome) Unsupervised learning: There is no structured label. Reinforcement Learning: we use a trial and error technique
  4. Artificial Neural Networks An interconnected group of nodes inspired the

    biological brain. An artificial neuron can receive a signal process it and then signal additional artificial neurons connected to it.
  5. Deep Learning is the subset of machine learning that has

    networks capable of learning unsupervised from unstructured and unlabeled data Deep learning is neural network of many layers It needs a lot of computing power (GPUs or TPUs)
  6. Artificial Neural Networks & Deep Learning Convolutional Neural Network are

    used to classify images Facebook adopts CNN for photo tagging
  7. CNN

  8. Note: Deep Learning needs a good training dataset to classify

    well. IMAGE CLASSIFICATION EXAMPLE Google iO|18 Diagnosing Diabetic Retinopathy
  9. Take a break! Have a chit-chat with your neighbor Touch

    up your makeup We will be back in 5
  10. Set-up our Environment • GPUs Vs CPUs • Jupyter Notebook

    • Keras, Tensorflow • Data Sets: Kaggle.com • Python
  11. GPU platforms you can use • Paperspace • Crestle •

    AWS • Google Cloud Platform • Azure Cloud