Lectures / Notes:

Below is the (tentative) list of classes, with possible additional readings. These may change as the semester progresses.
Date Topic (with linked notes / slides) Required readings Assignments / Deliverables Additional readings
23/08/2018 Introduction
28/08/2018 Image formation-geometry
[ ppt | pdf ]
Szeliski 2.1
30/08/2018 Image formation-color
[ ppt | pdf ]
Szeliski 2.2
4/9/18 Basic image processing
[ ppt | pdf ]
Szeliski 3.1-3.5, 4.2-4.3
6/9/18 Grouping
[ ppt | pdf ]
Contour Detection and Hierarchical Image Segmentation Szeliski 4.2-4.3, 5.2-5.3
11/9/18 Guest lecture 1
13/09/2018 Guest lecture 2
18/09/2018 Reconstruction - I
[ ppt | pdf ]
Project proposal due Szeliski 6
20/09/2018 Reconstruction - II
[ ppt | pdf ]
Space carving Szeliski 7
25/09/2018 Reconstruction - III | The correspondence problem
[ ppt | pdf ]
Assignment 1 Out Szeliski 7, 4
27/09/2018 The correspondence problem
[ ppt | pdf ]
SIFT
2/10/18 Optical flow
[ ppt | pdf ]
Szeliski 8.4
High accuracy optical flow estimation based on a theory of warping
4/10/18 Intro to ML
[ ppt | pdf ]
LeCun '98 (Till Section IV) Elements of Statistical Learning, Chapter 2
9/10/18
11/10/18 Neural networks - I
[ ppt | pdf ]
16/10/2018 Neural networks - II / Transfer Learning
[ ppt | pdf ]
Residual networks Backprop for convolution
Original ImageNet winner
Batch normalization
One of many papers
18/10/2018 Semantic segmentation
[ ppt | pdf ]
23/10/2018 Graphical models for refinement
[ ppt | pdf ]
Szeliski 3.7
Dense CRF
25/10/2018 Image generation
[ ppt | pdf ]
GAN Learning what and where to draw
30/10/2018 Object detection
[ ppt | pdf ]
R-CNN Assignment 2 out Fast R-CNN
Faster R-CNN
SSD
1/11/2018 Instance segmentation
[ ppt | pdf ]
Instance-FCN
Mask R-CNN
6/11/18 Pose estimation
[ ppt | pdf ]
Convolutional pose machines Autocontext
Inference machines
8/11/18 Structured prediction
[ pdf]
13/11/18 Learning 3D | Correspondence
[ ppt | pdf ]
Deep stereo Perspective transformer nets
15/11/2018 Correspondence | Vision and languag
[ ppt | pdf ]e
Visual QA
Modular architectures
Learning correspondence
20/11/2018 Embodied cognition
[ ppt | pdf ]
RL for navigation
22/11/2018
27/11/2018 Weak, semi, self-supervised and few-shot learning" Learning features by watching objects move
Prototypical networks
29/11/2018 Questions of ethics and bias Women also snowboard Gender shades
Paper on identifying sexuality with deep networks and response by Meg Mitchell and colleagues
4/12/18 Student Presentations All assignments due
15/12/2018 Final report due