Deep
Learning and Computer Vision (ECS795P)
Lectures: Wednesdays 15:00-17:00, Location: Queens Building QB-203
Labs:
Thursdays 13:00-15:00, Location: Queens Building QB-202, PC Room, 2nd
Floor
Lecturer: Sean Gong (web - www.eecs.qmul.ac.uk/~sgg/, email - s.gong@qmul.ac.uk, office - CS 325)
Course demonstrator: Jian Hu (web - https://lwpyh.github.io, email - jian.hu@qmul.ac.uk, office - CS 329)
Course demonstrator: Jiayi Lin (web - https://jylin8100.github.io/, email - jiayi.lin@qmul.ac.uk, office - C S329)
Course demonstrator: Zixu Cheng (web - https://zxccade.github.io/, email - zixu.cheng@qmul.ac.uk, office - CS 329)
Course demonstrator: Chaoran Zhu (web - https://noone65536.github.io/, email -
chaoran.zhu@qmul.ac.uk, office - CS
440)
Course demonstrator: Zengqun Zhao
(web - https://eecs.qmul.ac.uk/~zz012/,
email - zengqun.zhao@qmul.ac.uk,
office - Eng 152)
Course demonstrator: Yu Cao (web - https://www.yucao16.com/, email - yu.cao@qmul.ac.uk, office - CS 329)
Course
demonstrator: James Oldfield (web - https://eecs.qmul.ac.uk/~jo001/,
email - j.a.oldfiel@qmul.ac.uk,
office - Eng 152)
On Computer Vision and Machine
Learning
Chris Bishop, Pattern Recognition and Machine Learning (amazon)
- A benchmark
Simon Prince, Computer Vision (amazon)
- Intuitive
Kevin Murphy, Machine Learning: A Probabilistic Perspective (amazon)
- Comprehensive
David Marr, Vision (amazon)
- The Classic (there is more to deep learning)
Berthold Horn, Robot Vision (amazon)
- A Classic (when you really want to know)
Gene
Golub, Matrix Computations (amazon)
- A Classic on vectors, matrices, tensors
On Deep Learning
Ian Goodfellow, Deep Learning (amazon)
Richard Sutton, Reinforcement Learning (amazon)
W1 - Introduction to PyTorch by Facebook AI Research (FAIR)
Introduction to Deep Learning
& Computer Vision
W2 - What is deep learning (Chatfield et al BMVC'14, Zeiler et al
ECCV'14)
Convolutional Neural Networks (CNN)
CW1 Transformer handout (using Python in PyTorch)
W3 - Introduction to Transformers (Vasvani et
al NIPS'17, Dosovitskiy et al ICLR'21)
Transformer Critical Analysis Introduction & PyTorch Tutorial
W4 - Transformers deeper dive (Vasvani et al
NIPS'17, Dosovitskiy et al ICLR'21)
Questions & Answers on Transformers and PyTorch
Submit individual report on Transformers
W5 - Key concepts in CNN and the AlexNet (Krizhevsky et al NIPS'12, Srivastava et
al JMLR'14)
CW1 handin deadline (a written report
and code)
W6 - Computer vision with CNN and domain transfer learning (Girshick et al CVPR'14, Chen et
al CVPR'15)
CW2 Diffusion Model handout (using Python in PyTorch, teams of three students)
W7 - Introduction to Diffusion Models (Ho et al NIPS'20, Rombach et al
CVPR'22)
Diffusion Model Critical Analysis Introduction & Feedback
on Transformers report
W8 - Diffusion Model deeper dive (Ho et al NIPS'20, Rombach et al CVPR'22)
Questions & Answers on Diffusion Model and PyTorch
Submit individual report on Diffusion Model
W9 - CNN model for object detection - R-CNN/Fast R-CNN (Girshick ICCV'15, Uijlings et al
IJCV'13)
CW2 handin deadline (team reports & code,
individual self-reflection reports)
W10 - R-CNN model for action recognition (Gkioxari
et al ICCV'15)
CW1 online assessment individually
W11 - Going deeper: NIN, VGG, GoogLeNet (Lin et
al ICLR'14, Simonyan et al ICLR'15, Szegedy et al CVPR'15)
Feedback on Diffusion Model report (QMPlus)
W12 - Revision / CW3 mini-project specification and assessment criteria
handout
1. Course Works (30%)
(Overall
PyTorch code available + datasets; the students
are required to write certain modules in Python to complete the processing
pipelines, display/visualise/discuss results and
insights, together with a short written report - NO more than 4 pages A4)
Coursework Submission Guideline: (Please ensure that you
submit each of your completed coursework following the guidelines below. Failing
to comply may result in your coursework not to be marked with zero credit
for the specific coursework):
Submit each coursework through the QMPlus
online course website
Submit each coursework as a single zip file containing both code
and report
Your reports must be submitted as PDF files
Your code/implementation must be submitted as "XX.py" files
Each
submitted coursework zip file needs be named as "student_number.zip"
Coursework 1: Transformers (100 marks
for 15% of overall assessment)
CW1 handout Wk2 (Thursday 1 February)
CW1 submission DEADLINE Wk5 (23:55 Thursday 22 February)
CW1 online assessment, Wk10 (Thursday 28 March)
CW1 marks Wk12 (Thursday 11 April)
CW1 material - guideline, template, supporting code, and data
(download available HERE from Thursday 1 February)
(download CW1 paper from here)
Coursework 2: Diffusion Model (100 marks for 15% of overall assessment)
CW2 handout Wk6 (Thursday 29 February)
CW2 submission DEADLINE Wk9 (23:55 Thursday 21 March)
CW2 marks Wk12 (Thursday 11 April)
CW2 material - guideline, template, and supporting code
(download available HERE from Thursday 29 February)
(download CW2 paper from here)
P1 - Transformers (100 marks for 10% of the overall assessment)
Individual report submission DEADLINE Wk4 23:55 Thursday 15
Feb (submit through the QMPlus online in a zip
file)
Report marks Week 7
(download discussion paper P1 here, critical analysis report guideline here)
P2 - Diffusion Model (100 marks for 10% of the overall assessment)
Individual report submission DEADLINE Wk8 23:55 Thursday 14
Mar (submit through the QMPlus online in a zip
file)
Report marks Week 11
(download discussion paper P2 here,
individual report guideline here)
An individually assessed mini-project based on each student to submit
(1) Deep learning experimental evaluation test results, visualization,
Python code for the models and a written report;
(2) The report is a 6-pages research paper including the test results and visualisation plus any additional pages for references and extra supporting figures in an Appendix;
(3) CW3 guideline and assessment criteria (download available HERE from Thursday 4 April)
This assessment is to be completed individually by each student with a DEADLINE on Thursday 9 May 23:55.