I am an AI Research Scientist at Meta AI London, where I work on video diffusion models, generative rendering, and talking avatars.

I received my Ph.D. in Computer Science from the University of Surrey, supervised by Prof. Tao Xiang and Prof. Yi-Zhe Song, and worked closely with Dr. Xiatian (Eddy) Zhu.

I completed my Bachelor’s degree in Computer Science (Artificial Intelligence) at the University of Malaya (UM), where I was a lab member of CISIP under Prof. Chan Chee Seng.

You can contact me via kamwoh [at] gmail.com. Please see my CV at here.

Selected Publications

Generative Rendering

1. Kaleido (ICLR 2026) - A family of spatial generative models that achieves photorealistic, unified object- and scene-level neural rendering.

Multimodal Generation

1. VecGlypher (CVPR 2026) - A unified multimodal language model that generates high-fidelity, editable vector glyphs directly from text descriptions or image exemplars, bypassing raster intermediates.

Diffusion Models

1. PartCraft (ECCV 2024) - Crafting object by parts with text-to-image diffusion models.
2. Chirpy3D - Learning 3D fine-grained object generation from 2D unposed images.

Interpretable Fine-grained Hashing

1. ConceptHash (CVPRW 2024 FGVC11 Best paper award) - Learning to encode parts into interpretable codes.

Deep Hashing for Image Retrieval

1. DPN (IJCAI 2020) - Bit-wise hinge loss with random orthogonal codebook.
2. OrthoHash (NeurIPS 2021) - Cross entropy loss with random orthogonal codebook.
3. FIRe - Open source Fast Image Retrieval framework.
4. SDC (BMVC 2023 Oral) - Unsupervised hashing by maximizing the utilization of hash spaces through Wasserstein distance.

Deep Watermarking

1. DeepIPR (NeurIPS 2019) - Watermarking our DNNs by embedding “passport” and “signature”.
2. IPR-GAN (CVPR 2022) - Protecting our GANs by generating watermarked images.