Kevin Li is a post-doc researcher at SAYED Systems Group, working with Dr. Ahmed Sayed. He has a Ph.D. in Computer Science from Queen Mary University of London, supervised by Prof. Shaogang (Sean) Gong FREng. He previously earned an M.S. degree from Sichuan University in 2020. From June 2022 to April 2024, he worked as a machine learning scientist at Veritone Inc, where he focused on developing a scalable person search framework for retrieving individuals at different locations and times, as captured by various cameras. His current research interests lie in privacy-aware distributed machine learning, with a particular emphasis on learning domain-invariant knowledge representation from multimodal data captured in diverse environments. His research outcome has been recognized as ESI Highly Cited Paper (Top 1%). Additionally, he serves as an evaluator for the ELLIS PhD Program.
Selected Publications [ Full List ]
-
Feature-Distribution Perturbation and Aalibration for Generalized Person
ReID
Qilei Li, Jiabo Huang, Jian Hu, Shaogang Gong
In Proc. IEEE International Conference on Acoustics, Speech, and Signal Processing, (ICASSP), Korea, April 2024, Oral, New
[ PDF ] [ Link ] -
Mitigate Domain Shift by Primary-Auxiliary Objectives Association for
Generalizing Person ReID
Qilei Li, Shaogang Gong
In Proc. IEEE/CVF Winter Conference on Applications of Computer Vision, (WACV), United States, January 2024, New
[ PDF ] [ Link ] -
Towards multimodal disinformation detection by Vision-language Knowledge
Interaction
Qilei Li, Mingliang Gao, Guisheng Zhang, Wenzhe Zhai, Jinyong Chen, Gwanggil Jeon
Information Fusion, September 2023, IF=14.7
[ PDF ] [ Link ] -
Efficient Pansharpening by Joint-modality Recursive Training
Qilei Li†, Wenhao Song†, Mingliang Gao, Wenzhe Zhai, Jianhao Sun, Gwanggil Jeon
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, July 2024
[ PDF ] [ Link ] ] -
Defending Deepfakes by Saliency-Aware Attack
Qilei Li, Mingliang Gao, Guisheng Zhang, Wenzhe Zhai
IEEE Transactions on Computational Social Systems, May 2023
[ PDF ] [ Link ]
Workshop
-
Image processing and deep learning Technologies in
Communication,
International Conference on 6G Communication and Internet of Things Technology (6GIoTT), Fuzhou, China, Oct. 2022
-
Computer Vision Technologies for Smart City
Applications,
Frontiers of Electronics, Information and Computation Technologies (ICFEICT), Yangzhou, China, Apr. 2024
-
Advancing AIGC: Faithful Generation and
Trustworthy Identification,
International conference on Computer, Big Data and Artificial Intelligence (ICCBD+AI), Jingdezhen, China, Nov. 2024
Work Experience
-
Research Assistant (Full-time)
National University of Singapore (Singapore), Aug. 2020 – Oct. 2020
Worked on real-world image super-resolution with deep learning. -
Deep Learning Algorithm Engineer (Full-time)
Vision Semantics Limited (London, UK), July 2022 – Oct. 2022,
Improved person ReID framework using foundation models (ViT, CLIP, BLIP) for accurate retrieval of pedestrians. -
Analysis Algorithm Engineer (Part-time, 40% Full time)
Veritone Limited (London branch, US), Oct. 2022 – April 2024
Implementing large-scale models for object detection, applying domain generalization techniques to enhance real-world performance.
Patents
-
2020.08.03 - A sparse coding based infrared and visible image fusion method
First applicant, Substantive examination stage, CN114066786A
The invention discloses a fusion method of infrared image and visible image, which produces excellent results with rich details and gradient information. -
2019.09.16 - An image super-resolution reconstruction method based on gated multi-feedback
networks
First applicant, Substantive examination stage, CN112508779A
The invention presents an image super-resolution reconstruction method using gated multi-feedback networks to enhance image quality. -
2019.09.12 - Infrared and visible image fusion method based on coupling generation
adversarial network
First applicant, Substantive examination stage, CN112488970A
The invention introduces an infrared and visible image fusion method using a coupling generation adversarial network to maintain thermal radiation and texture information effectively. -
2019.09.16 - A multi-focus image fusion method based on sparse representation and guided
filtering
Second applicant, Substantive examination stage, CN112508828A
The invention discloses a multi-focus image fusion method, which effectively fuses pictures with different focuses, displaying more information while maintaining sharpness and resolution.