Publications

You can also find my articles in my Google Scholar. (* indicates equal contribution. )

2022

CyEDA: Cycle-Object Edge Consistency Domain Adaptation [ICIP] [PDF] [Code]
We proposed blending masks and cycle-object edge consistency to preserve instance-level details in an image during day-night translation.
Jing Chong Beh, Kam Woh Ng, Jie Long Kew, Che-Tsung Lin, Chee Seng Chan, Shang-Hong Lai, Christopher Zach. In ICIP, 2022.

Large-Scale Product Retrieval with Weakly Supervised Representation Learning [arXiv] [Code]
We mine pseudo-attributes from product titles for weakly supervised representation learning.
Xiao Han*, Kam Woh Ng*, Sauradip Nag, Zhiyu Qu. In FGVC9 CVPR, 2022.

2021

One Loss for All: Deep Hashing with a Single Cosine Similarity based Learning Objective [NeurIPS] [Code]
We proposed a simple and elegant deep hashing method with only a single classification learing objective.
Jiun Tian Hoe*, Kam Woh Ng*, Tianyu Zhang, Chee Seng Chan, Yi-Zhe Song, Tao Xiang. In NeurIPS, 2021.

DeepIP: Deep Neural Network Intellectual Property Protection with Passports [TPAMI] [Code]
We further extend the work of ownership verification with “passport” by completely describing how we perform ambiguity attack on a protected model.
Lixin Fan*, Kam Woh Ng*, Chee Seng Chan, Qiang Yang. In TPAMI, 2021.

Protecting Intellectual Property of Generative Adversarial Networks from Ambiguity Attack [CVPR] [Code]
We proposed an intellectual property right protection framework for Generative Adversarial Networks.
Ding Sheng Ong, Chee Seng Chan, Kam Woh Ng, Lixin Fan, Qiang Yang. In CVPR, 2021.

2020

Rethinking Privacy Preserving Deep Learning: How to Evaluate and Thwart Privacy Attacks [Report] [Book]
We proposed Secret Polarization Network to perturb exchanged gradient in order to defend against Deep Leakage attack.
Lixin Fan*, Kam Woh Ng*, Ce Ju*, Tianyu Zhang, Chang Liu, Chee Seng Chan, Qiang Yang.

Deep Polarized Network for Supervised Learning of Accurate Binary Hashing Codes [IJCAI] [Code]
We proposed polarization loss, a differentiable hinge-like loss, for learning accurate hash codes.
Lixin Fan*, Kam Woh Ng*, Ce Ju, Tianyu Zhang and Chee Seng Chan. In the 29th International Joint Conference on Artificial Intelligence (IJCAI), 2020.

2019

Rethinking deep neural network ownership verification: Embedding passports to defeat ambiguity attacks [NeurIPS] [Code] [Website]
We proposed embedding “passport” into deep neural network for ownership verification.
Lixin Fan*, Kam Woh Ng*, Chee Seng Chan. In 33th Conference on Neural Information Processing Systems (NeurIPS), 2019.

A Universal Logic Operator for Interpretable Deep Convolution Networks [arXiv]
We proposed to apply fuzzy logic in Deep Convolutional Network for interpretability.
Kam Woh Ng*, Lixin Fan*, Chee Seng Chan. In AAAI-19 Workshop on Network Interpretability for Deep Learning.