CS代考 CSCI435/CSCI935

CSCI435/CSCI935
Computer Vision – Algorithms and Systems
Subject Review & Final Exam
Lecturer: Assoc/Prof Wanqing Li
Room 3.101
Web: http://www.uow.edu.au/~wanqing
25/10/2021

Subject Learning Outcomes
On successful completion of this subject, students are expected to:
• Understand the principle of digital image and video cameras.
• Use image enhancement techniques.
• Use object detection and recognition techniques.
• Use video processing techniques to detect moving
• Design and implement basic computer vision systems
for real applications.
25/10/2021 2

Topics Covered in the Subject
 Photometry and colourimetry
 light, colour perception and colour spaces
 Image acquisition
 Optical system. sampling, image sensors, single sensor based
digital camera, colour processing chain  Image quality & enhancement
 Criteria of quality, sharpness, low- & high-pass filter in spatial and frequency domain, enhancement, noise, image spectrum and pyramids
 Edge detection
 Gradient, edge detection operators, zero-crossing, LoG, DoG,
Canny edge detector  Key point detection
 Harris corner detection, SIFT interest points and descriptors, BoW, image similarity
25/10/2021 3

Topics Covered in the Subject
 Shape detection
 Hough transform (line), circle detection
 Image segmentation
 Visual features, perceptual grouping, thresholding (heuristic &
Otsu’s), clustering-based (k-means, mean-shift)  Binary image processing
 Binary morphology, connected component analysis  CD and background modelling
 Robust CD, Background modelling (running average/median/Gaussian GMM)
 Object detection
 General framework (detection as classification), sliding window vs. reginal proposal (selective search), skin-colour based face detection, AdaBoost (Viola & Jones detector), HoG for detection of humen and faces
25/10/2021 4

Topics Covered in the Subject
 Image classification and object recognition
 General framework, human perception of faces, face recognition system, normalization of faces, eigenfaces, LBP-based face recognition
 Motion estimation
 Optical flow, HS method, LK method, global motion, motion
analysis and its applications
 Convolitional Neural Networks (ConvNets)
 Linear classifier, softmax classifiers, optimization, multiple layer perceptron (fully connected layers), gradient backpropagation, convolutional layers, learning ConvNet parameters (mini-batch SGD, batch normalization), hyper-parameters, regularization and dropout, data augmentation, typical ConvNets for CV
25/10/2021 5

Subject Materials for Review
Lecture slides:
 Available on the subject Moodle.
Recommended books:
 D. Forsyth, J. Ponce. Computer Vision a Modern Approach,
, 2012 (2nd ed.)
 E. R Davies, Computer and machine vision: theory, algorithms and practicalities, Academic Press; 4th edition; 2012
 Stanford’s course Convolutional Neural Networks for Visual Recognition http://cs231n.stanford.edu/
Assignments
25/10/2021 6

Assessments
Assignments (60% in total)
 3x Coding projects 3 = 60%
Final Exam (40%)
 Minimum requirement 40% = 16 marks
25/10/2021 7

Final Examination
Materials and Aids Allowed  Open book
Exam Structure
 Problem solving and discussion
 4 questions, 10% each
 Each question has multiple sub-questions
This exam will run via Moodle
25/10/2021 8

Final Examination…
Exam Date & Starting time
 13:30 (Sydney time) Monday 15 November
 Please check SOLS
Exam Duration  2 hours
• 30 minutes for preparing and submitting answer sheets in a
single pdf file
25/10/2021 9

Final Examination – Instructions
Have a set of A4 blank paper ready On the first page, write
 Your full name, Student Number & UOW login name Answer each question on a separate page clearly
 either handwriting or using suitable editing software at your own choice
Scan or take photos of your answer sheets and convert them into one single pdf file (<200MB) Name the pdf file as  .pdf
Submit the pdf file via Moodle
 See the next slide on how to scan/convert your hand-write answer sheets into a
single pdf file using your mobile
25/10/2021 10

How to create one pdf file
 Important: Be prepared with knowing how to create one pdf file from your working solutions.
 There is freely available software that can be used to scan your answer sheets and convert them into a single pdf file. These links may be of assistance.
https://www.youtube.com/watch?v=BCccqxhPyJw (Scan documents) https://www.youtube.com/watch?v=d_olWftSgIM (Convert image to pdf)
 iPhone https://www.idownloadblog.com/2017/05/12/how-to-save-photos-pdf-
iphone-ipad/
https://www.igeeksblog.com/how-to-convert-photos-to-pdf-on-iphone- ipad/
25/10/2021 11

Example Problems
Disclaimer
This is not an exclusive list of problems that may appear in the final exam, they are just examples
25/10/2021 12

Example Problems
Single sensor based cameras and image processing
 Key components
 How each component affects quality of
 Noise propagation
 How to enhance images with low visual quality
25/10/2021 13

Example Problems
Automatic Recognition of the following road sign in images
Automatic counting the number of balls
25/10/2021 14

Example Problems
People Counting
Detection of car registration Classification of vehicles Detection of hands Problems in assignments
25/10/2021 15

Types of Possible Questions
 How will you classify this problem with regards to computer vision problems you have studied in the class?
 Propose a solution to the problem. Divide the solution into components and describe the solution using a block diagram or flowchart. Explain the function, input and output of each components.
 For each component in the solution, choose suitable algorithms and briefly describe how the algorithms works.
 Describe how you would test your solution and measure its performance.
 Discuss whether your algorithm would work in “certain” conditions, Explain why it works or why it does not work.
 What are the possible factors that may affect the accuracy of your system?
25/10/2021 16

How to contact me
Consultation via Zoom
 Monday 15:30 – 17:30  Wednesday 16:30 – 18:30
 Set the subject of the email as
o CSCI435 or CSCI935: (topic of the email)  Will be responded as soon as possible
25/10/2021 17

25/10/2021 18