​

Part 1: Introduction and Tutorial Infrastructure

​

We will review the outline of the course and ensure all attendees have access to the course slides and can install all necessary tools. e.g. jupyter notebook, python libraries, etc.

30 min

 

Part 2: Traditional Computer Vision

In this session, we will cover traditional computer vision techniques including Haar, SIFT, and HOG features.

We will also introduce the OpenCV frameworks as well as Viola-Jones facial detection.

30 min

 

Break 15 min

 

Part 3: Convolutional Neural Networks

Review the most common operations of CNN Architectures, pooling, convolutions and fully connected layers.

This session will then cover popular CNNs used in the industry including VGGNet, Inception, and Resnet.

75 min

 

Lunch Break 60 min

 

Part 4: Object Detection

This session will introduce object detection with popular two-stage approaches: R-CNN, Fast R-CNN, and Faster R-CNN.

We will then cover faster single stage approaches including SSD and YOLO.

75 min

 

Break 15 min

 

Part 5: Object Segmentation

We will discuss the most popular segmentation approaches using fully convolutional networks: FCN, U-Net, Mask-RCNN and PSPNet.

75 min

 

Break 15 min

 

Part 6: Advanced applications

This final session will discuss generative approaches by using deep learning to generate synthetic images with GANs.

This will cover popular networks such as CycleGAN and Pix2pix.

60 min