|
Course description. Computer Vision is broadly
defined as the study of recovering useful properties of the world
from one or more images. In recent years, Deep Learning has
emerged as a powerful tool for addressing computer vision
tasks. This course will cover a range of foundational topics
at the intersection of Deep Learning and Computer Vision.
Contact information.
kosta[at]cs.ryerson[dot]ca
Prerequisites. University-level courses on
linear algebra, multivariable calculus, and probability
Required textbook. Ian Goodfellow, Yoshua Bengio
and Aaron Courville, Deep Learning (available for free or
purchase)
Optional textbook. Marc Peter Deisenroth, A Aldo Faisal, and Cheng Soon Ong, Mathematics for Machine Learning, (preprint available for free)
Lectures.
# | TOPIC | SLIDES ![]() ![]() |
---|---|---|
0 |
Introduction to Computer Vision | PDF MOV |
1 |
Ethics,
Privacy and Security in Machine Learning |
PDF MOV |
2 |
Machine Learning Crash Course | PDF MOV |
NEURAL NETWORK
FOUNDATIONS |
||
3 |
Multilayer
Perceptron |
PDF MOV |
4 |
Convolutional
Network (ConvNet) |
PDF MOV |
5 |
Backprop |
PDF MOV |
6 |
Optimization |
PDF MOV |
7 |
Introduction
to PyTorch (version 1.0) |
PDF MOV |
SPATIAL MODELS |
||
8 |
Object Recognition Architectures | PDF MOV |
9 |
Training
Networks |
PDF MOV |
10 |
Weight
Initialization |
PDF MOV |
11 |
Transfer Learning | PDF MOV |
12 |
Object Detection | PDF MOV |
13 |
Metric
Learning |
PDF MOV |
14 |
Pixel Labeling Tasks | PDF MOV |
15 |
Optical
Flow |
PDF MOV |
16 |
Segmentation
Aware Filtering |
PDF MOV |
VISUALIZATION |
||
17 |
Understanding ConvNets |
PDF MOV |
18 |
Texture
Synthesis |
PDF MOV |
19 |
Style
Transfer |
PDF MOV |
SEQUENTIAL MODELS | ||
20 |
Recurrent Neural Network (RNN) |
PDF MOV |
21 |
Long Short-Term Memory (LSTM) |
PDF MOV |
22 |
Bidirectional
RNN |
PDF MOV |
23 |
Vision
and Language |
PDF MOV |
24 |
Action Recognition |
PDF MOV |
SELF-SUPERVISED LEARNING |
||
25 |
Look ma no labels: Learning without
labels |
PDF MOV |
GENERATIVE
MODELS |
||
26 |
PixelNN (PixelRNN and
PixelCNN) |
PDF MOV |
27 |
Variational Autoencoder (VAE) |
|
28 |
Invertible Density
Models - Normalizing Flows |
|
29 |
Generative
Adversarial Network (GAN) |
|
ADVERSARIAL EXAMPLES | ||
30 |
Adversarial
Examples |
PDF MOV |
Related courses.
Ryerson
University CPS843: Introduction
to Computer Vision with Kosta Derpanis
Stanford CS231N: Convolutional Neural Networks for
Visual Recognition with Fei Fei Li, Justin Johnson and Serena Young
EPFL EE-559: Deep Learning with Francois Fleuret
Acknowledgements. While a great effort has been made to
assemble an original set of lecture slides, the essence of the
presentation of many of the slides rely on material prepared
by the following people: Andrej
Karpathy, Justin Johnson, Serena Young, Fei Fei Li,
Francois Fleuret, Graham Taylor,
Carl Doersch