Computer Vision
CSE P576 // Spring 2018
Lectures: Thursdays 6:30 - 9:20 pm, Johnson Hall (JHN) 075
Office Hours: Wednesdays 5:00 - 6:00 pm, CSE 674, or by appointment
Matthew Brown, Richard Newcombe, Rob Gens, David Rosen, Steven Lovegrove, Jonathan Huang, Jakob Engel, Yuting Ye
TAs: Tanner Schmidt (tws10@cs.washington.edu), Melody Su (yhsu83@uw.edu)
A professional masters course in computer vision, emphasizing fundamentals of geometry and image formation as well as deep learning and image understanding. Covers low-level image processing, image geometry, motion and depth estimation, object recognition, deep learning and optimization techniques, and case studies of current research.
Grading:The grade is based on four projects. Each project will be a mix of coding and written answers. See the course overview below for handin dates. Late policy is as follows: Late assignments will be accepted up to 5 days after the deadline, but with 10% subtracted from the mark per day.
| Week | Date | Topic | Description |
|---|---|---|---|
| 1 | 3/29 | Introduction | |
| Image Formation | Geometric and Photometric Image Formation, Pinhole Camera, Lenses, Sensors, Colour, Gamma, DCT, Image Coding | ||
| Week 1 Notes | |||
| 2 | 4/5 | Filtering and Pyramids | Linear + Non-linear filtering, Correlation, Convolution, Gaussian + Laplacian Pyramids, Sampling and Aliasing |
| Features and Matching | Detection, Correspondence, Edges, Corners, Regions, Patch matching, SIFT, Shape Context, Learning Features | ||
| Week 2 Notes | |||
| Project 1 start | |||
| 3 | 4/12 | Planar Geometry | Image Alignment, Linear and Projective Cameras + 2D Transforms |
| RANSAC | Robust 2-view geometry estimation | ||
| Epipolar Geometry | Epipolar Lines, Plane Constraint, Fundamental Matrix, Stereo Matching | ||
| Week 3 Notes | |||
| 4 | 4/19 | Multiview Geometry | SFM/SLAM, 3D points, Camera Pose and Calibration, Bundle Adjustment |
| Optimization | Noise Modelling, Non-L2 residuals, Convexity, Problem Solving | ||
| Project 2 start | |||
| 4/22 | Project 1 due | ||
| 5 | 4/26 | Camera Tracking | Sparse vs dense, brightness constancy, optical flow, coarse to fine, lucas kanade, incremental rotation, dense tracking + mapping, ICP |
| Dense Reconstruction | Stereo, disparity space, aggregation/block matching, plane sweep, baseline, denoising, depth priors, signed distance functions + depth fusion, voxel colouring | ||
| LucasKanade ipynb | |||
| PlaneSweep ipynb | |||
| 6 | 5/3 | Machine Learning for Vision | Intro, inductive learning, decision trees, instance-based learning |
| Project 3 start (ipynb, html) | |||
| 5/6 | Project 2 due | ||
| 7 | 5/10 | Convolutional Neural Networks | Neural Networks, Backpropagation, Convolutional Neural Nets, Training Neural Networks, Additional slides |
| 8 | 5/17 | Object Detection | Sliding windows, detection with convnets, anchor regression+classification, SSD, Faster R-CNN, R-FCN, evaluation, IOU, AP, instance segmentation, keypoint detection |
| Depth Estimation | Unsupervised Monocular Depth [Godard et al.] | ||
| Project 4 start | |||
| 5/20 | Project 3 due | ||
| 9 | 5/24 | Hand Tracking | |
| Dense Reconstruction | |||
| SLAM | SLAM+VO: (Jakob Engel) Direct/indirect, sparse/dense, DSO, ORB-SLAM, photometric calibration, practical tips, evaluation, analysis | ||
| 10 | 5/31 | History and Futurology | |
| 6/3 | Project 4 due |