menu Computer Vision/AI using Deep Learning
Left Brain





Objective

This class is intended for people who want to get a practical understanding on how to solve computer vision problems using deep learning / tensorFlow. In this course we will discuss some of the most widely used deep learning algorithms applied to solve computer vision problems. The students will build models using convolutional neural networks to solve computer vision problems. The application of the models will be in image classification, image segmentation, object detection, and image extraction. TensorFlow is an open-source machine learning library for research and production. The students will use python and tensorflow APIs to construct and train models in this course. Students will experience creating supervised and unsupervised learning tensorFlow models.

    Agenda

    - Introduction to Computer Vision
      a. Image processing and computer vision
      b. Using deep learning
    - Convolution Neural Networks
      a. Convolution layer, pooling layer, upsampling layer
      b. Architectures discussion: AlexNet, VGG, GoogLeNet, ResNet and much more
    - Unsupervised Deep Learning – AutoEncoders
      a. Overview, use of autoencoders in computer vision
      b. Simple encoder and decoder, sparse autoencoder, variational autoencoder, and denoising autoencoder
    - Image segmentation
      a. Image segmentation overview
      b. Preparing training data, loss functions
    - Object detection
      a. Overview, use cases
      b. RCNN / Fast RCNN / Faster RCNN, YOLO, SSD, and Masked RCNN
    - Considerations in implementation
      a. Optimize models for size, speed, and accuracy
    - Project

Intended Audience

Programmers, managers, investors, enthusiast pretty much anyone technically curious about deploying artificial intelligence. If you are familiar with supervised learning but did not know how to implement them in tensorflow.

What do I need to know to be successful:

All you need is curiosity in the subject and access to a laptop for which you have administrator access. (Specifically so you can turn off firewall protection in case you have issues connecting to the cloud). We will grant you access to a Google Cloud servers with a connected GPU for conducting the hands-on exercise. About 30% of the class is lectures and 70% is hands on programming. In order to do the hands-on programming segment you need to know python. A general understanding of numpy, scipy and tensorflow would help. But since people have varying experience with these topics, we will be providing a quick overview.




info@leftbrain.consulting
© 2018 Copyright: Left Brain LLC.