Infinite-ID: Identity-preserved Personalization via ID-semantics Decoupling Paradigm

Yi Wu, Ziqiang Li (Equal Contribution), Heliang Zheng, Chaoyue Wang, Bin Li.

Project

• An ID-semantics decoupling paradigm for tuning-free identity-preserved personalization.

Peer is Your Pillar: A Data-unbalanced Conditional GANs for Few-shot Image Generation

Ziqiang Li, Chaoyue Wang, Xue Rui, Chao Xue, Jiaxu Leng, ZhangJie Fu, Bin Li

• A novel pipeline called Peer is your Pillar (PIP), which combines a target few-shot dataset with a peer dataset to enable data-unbalanced conditional generation.

One-Shot Generative Domain Adaptation in 3D GANs

Ziqiang Li (Equal Contribution), Yi Wu, Chaoyue Wang, Xue Rui, Bin Li.

• Introducing a novel task, One-shot 3D Generative Domain Adaptation (GDA), which focuses on transferring a pre-trained 3D generator from one domain to a new one using only a single reference image.

A Proxy Attack-Free Strategy for Practically Improving the Poisoning Efficiency in Backdoor Attacks

Ziqiang Li, Hong Sun, Pengfei Xia, Beihao Xia, Xue Rui, Wei Zhang, Qinglang Guo, Zhangjie Fu, Bin Li.

• A Proxy attack-Free Strategy (PFS) designed to identify efficient poisoning samples based on the similarity between clean samples and their corresponding poisoning samples.

Large Language Models are Good Attackers: Efficient and Stealthy Textual Backdoor Attacks

Ziqiang Li, Yueqi Zeng, Pengfei Xia, Lei Liu, Zhangjie Fu, Bin Li.

• Leveraging Large Language Models (LLMs) to acquire Efficient and Stealthy Textual backdoor attack method.

Efficient Backdoor Attacks for Deep Neural Networks in Real-world Scenarios

Ziqiang Li (Equal Contribution), Hong Sun, Pengfei Xia, Heng Li, Beihao Xia, Yi Wu, Bin Li.

• Introducing a realistic backdoor attack scenario where victims collect data from multiple sources, and attackers cannot access the complete training data.

Domain Re-Modulation for Few-Shot Generative Domain Adaptation

Yi Wu, Ziqiang Li (Equal Contribution), Chaoyue Wang, Heliang Zheng, Shanshan Zhao, Bin Li, and Dacheng Tao.

• Not only fulfills three essential properties of few-shot GDA but also introduces two new capabilities: Memory and Domain Association.

A Systematic Survey of Regularization and Normalization in GANs

Ziqiang Li, Muhammad Usman, Rentuo Tao, Pengfei Xia, Chaoyue Wang, Huanhuan Chen, and Bin Li.

• Conducting a comprehensive survey on the regularization and normalization techniques from different perspectives of GANs training.

Exploring The Effect of High-frequency Components in GANs Training

Ziqiang Li, Pengfei Xia, Xue Rui, and Bin Li.

• Proposing two simple yet effective frequency operations for eliminating the side effects caused by high-frequency differences in GANs training.

A New Perspective on Stabilizing GANs Training: Direct Adversarial Training

Ziqiang Li, Pengfei Xia, Rentuo Tao, Hongjing Niu, and Bin Li.

• Images produced by the generator act like adversarial examples of the discriminator during the training process, which may be part of the reason causing the unstable training of GANs.