Kin Wai Lau

Kin Wai Lau

Research Engineer -- AI / Deep Learning / Machine Learning

Kin Wai Lau

I am a Research Engineer in the TCL AI Lab where I work on Computer Vision, Audio Processing, and Deep Learning. I received the B.E. degree in Information Engineering from City University of Hong Kong, in 2017. Over the past six years, I have gained valuable experience working on various projects, including audio recognition for smart home robots, anomaly detection for industrial applications, and low-level image enhancement tasks such as super-resolution and diffraction removal for under display cameras (UDC).

In addition to my full-time work as an engineer, I am currently pursuing a part-time PhD degree at Department of Electrical Engineering, City University of Hong Kong. During my PhD study, I am working under the guidance of Dr. Lai-Man Po on the development of efficient and lightweight networks for image and audio recognition utilizing techniques like reparameterization.

Research Interests

Computer Vision, Audio Processing, Image enhancement, Image and audio representation learning, Federated learning, Self-supervised learning, Reparamerterization.

News

  1. [September 2024] Our paper titled "FedRepOpt: Gradient Re-parametrized Optimizers in Federated Learning" has been accepted to the ACCV 2024. [Paper] [Code]

  2. [April 2024] Our paper titled "Exploring Federated Self-Supervised Learning for General Purpose Audio Understanding" has been accepted to the ICASSP-2024 workshop on Self-supervision in Audio, Speech and Beyond. [Preprint]

  3. [February 2024] Our paper titled "AudioRepInceptionNeXt: A lightweight single-stream architecture for efficient audio recognition" has been accepted in NeuroComputing. [Paper] [Code]

  4. [November 2023] Our paper titled "Adaptive uncertainty estimation via high-dimensional testing on latent representations" has been accepted in NeurIPS 2023. [Paper] [Code]

  5. [August 2023] Our paper titled "Large Separable Kernel Attention: Rethinking the Large Kernel Attention design in CNN" has been accepted in Expert Systems with Applications. [Paper] [Code]

  6. [June 2023] Our solution got first place award in EPIC-Kitchens-100 2023 Challenges on EPIC-SOUNDS Audio-Based Interaction Recognition, CVPR 2023 Workshop. [Website] [Report] [Code]

  7. [June 2022] Our paper titled "What Should Be Equivariant in Self-Supervised Learning" has been accepted to CVPR Workshops on L3D-IVU. [Paper]