News
2024/09/08: [Reminder: Classroom change] We now change to BL-113 to accommodate a total of 150 students, starting Sept. 10th.
2024/09/06: [Course registration update] The permission code for course registration has been sent out. If you do not plan to take the course, please let the instructor know ASAP so that we can have more people in.
2024/08/08: Please refer to the policy in Week 1 (W1-1) slides if you are NTU students and would like to register this course.
2024/09/06: [Course registration update] The permission code for course registration has been sent out. If you do not plan to take the course, please let the instructor know ASAP so that we can have more people in.
2024/08/08: Please refer to the policy in Week 1 (W1-1) slides if you are NTU students and would like to register this course.
Introduction
This course focuses on the transformative power of deep learning based approaches in computer vision. Students will learn fundamental knowledge and explore state-of-the-art techniques, including convolutional neural networks (CNNs), generative adversarial networks (GANs), Transformer, Diffusion Models and 3D vision, with a strong emphasis on model designs and their practical applications. Through rigorous hands-on projects and in-depth theoretical discussions, you will gain a comprehensive understanding of how deep learning models can be designed, trained, and optimized for complex visual tasks. This course prepares you for advanced research and professional roles in the ever-evolving field of computer vision.
Goals
This course will expose students to cutting-edge research — starting from fundamentals of deep learning to its recent advances in various vision applications. You will be expected to master key concepts, such as neural network architectures and their designs, training techniques, and performance optimization strategies. As a major part of this course, final projects are offered to foster critical thinking and problem-solving skills, preparing you to contribute to cutting-edge research or tackle real-world challenges in the real-world problems. Active participation, completion of hands-on projects, and engagement in theoretical discussions will be crucial to your success in this course.
Syllabus
Week |
Date |
Topic |
Course Materials |
Remarks |
1 |
09/03 |
Course Logistics & Registration; Intro to Neural Nets |
||
2 |
09/10 |
Convolutional Neural Networks & Image Segmentation |
W2 |
HW #1 out |
3 |
09/17 |
No class |
Mid-Autumn Festival |
|
4 |
09/24 |
Generative Models (I) - AE, VAE, & Diffusion Model (I) |
W4 |
HW #1 due |
5 |
10/01 |
Guest Lecture: Dr. Jun-Cheng Chen, Academia Sinica |
ECCV week |
|
6 |
10/8 |
Generative Models (II) - Diffusion Model (II), GAN |
HW # 2 out |
|
7 |
10/15 |
Recurrent Neural Networks & Transformer |
||
8 |
10/22 |
Transformer; Vision & Language Models |
||
9 |
10/29 |
Vision & Language Models; Multi-Modal Learning |
HW #2 due; HW #3 out |
|
10 |
11/05 |
Parameter-Efficient Finetuning; Unlearning, Debiasing, and Interoperability |
W10 |
|
11 |
11/12 |
Guest Lecture: Linda Huang, Senior Dir., GeValyn Associates Chenchi Kuo, Senior Dir., Marvell Technology |
||
12 |
11/19 |
3D Vision |
HW #3 due; HW #4 out |
|
13 |
11/26 |
Object Detection |
Final Project Announcement |
|
14 |
12/03 |
Guest Lecture: Prof. Ming-Ching Chang, SUNY, Albany, USA Guest Lecture: Dr. Trista Chen, Director, Microsoft AI Research Center, Taiwan |
HW #4 due |
|
15 |
12/10 |
Zero-Shot vs. Continual Learning with Foundation Models; Progress Check for Final Projects & General Q&A |
||
17 |
12/26 Thu |
Final Project Presentation |
Contacts
Teaching Assistants
Bin-Shih Wu
MK-514 TA Hours: Mon. 14:20 ~ 15:10 |
Jr-Jen Chen
MK-514 TA Hours: Fri. 14:20 ~ 15:10 |
Wei-Yuan Cheng
MK-514 TA Hours: Fri. 14:20 ~ 15:10 |
Yu-Hsiang Huang
MK-514 TA Hours: Mon. 14:20 ~ 15:10 |
Hung-Kai Chung
MK-514 TA Hours: Thu. 14:20 ~ 15:10 |
Yu-Ju Cheng
MK-514 TA Hours: Tue. 14:20 ~ 15:10 |
Hsin-Chen Lin
MK-514 TA Hours: Wed. 14:20 ~ 15:10 |
Fang-Duo Tsai
MK-514 TA Hours: Wed. 14:20 ~ 15:10 |