Publications

2025

1. AAAI’25

   

   ARNet: Self-Supervised FG-SBIR with Unified Sample Feature Alignment and Multi-Scale Token Recycling

   Jianan Jiang, Han Tang, Zhilin Jiang, Weiren Yu, and Di Wu

   In Proceedings of the AAAI Conference on Artificial Intelligence, 2025

   Abs DOI PDF Code

   We achieved better feature alignment and representation through a straightforward framework design by optimizing features for both intra- and inter-samples, and recycling discarded patch tokens in ViT-based backbones via contrastive learning.

2024

1. Preprint

   

   HAIFIT: Human-to-AI Fashion Image Translation

   Jianan Jiang, Xinglin Li, Weiren Yu, and Di Wu

   arXiv preprint arXiv:2403.08651, 2024

   Abs PDF Code

   We first treated input sketches as sequences to maintain consistency from forward and reverse directions, inspired by human perception.

2. IEEE T-II’24

   

   FPGA Adaptive Neural Network Quantization for Adversarial Image Attack Defense

   Yufeng Lu^*, Xiaokang Shi^*, Jianan Jiang, Hanhui Deng, Yanwen Wang, and 2 more authors

   IEEE Transactions on Industrial Informatics, 2024

   Abs DOI

   We achieved image adversarial attack and defense on the FPGA platform while ensuring model performance.

3. Ubicomp’24

   

   HAIGEN: Towards Human-AI Collaboration for Facilitating Creativity and Style Generation in Fashion Design

   Jianan Jiang, Di Wu, Hanhui Deng, Yidan Long, Wenyi Tang, and 5 more authors

   Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 2024

   Abs DOI PDF Video Code

   We introduced a system to facilitate the entire design process with a Human-AI collaborative framework, harnessing the powerful capabilities of cloud large models and the convenience of local small models through well-designed methodologies for each modules.

4. CSCW’24

   

   StyleWe: Towards Style Fusion in Generative Fashion Design with Efficient Federated AI

   Di Wu^*, Mingzhu Wu^*, Yeye Li^*, Jianan Jiang, Xinglin Li, and 3 more authors

   Proceedings of the ACM on Human-Computer Interaction, 2024

   Abs DOI Code Best Paper Award

   One of 5 Awardees, 1%.

   We introduce StyleWe, which employs a lightweight GAN with model compression and a federated learning algorithm to generate unexpected sketch styles by facilitating sketch styles fusion among multiple designers while ensuring privacy protection.

5. Ubicomp’24

   

   CrossGAI: A Cross-Device Generative AI Framework for Collaborative Fashion Design

   Hanhui Deng^*, Jianan Jiang^*, Zhiwang Yu, Jinhui Ouyang, and Di Wu

   Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 2024

   Abs DOI Video

   We introduce CrossGAI, which employs multiple AI modules and the Lyapunov algorithm with a DNN actor to dynamically optimize network bandwidth across different clients, supporting collaboration among multiple designers on various devices across different regions.

2023

1. CHI’23

   

   Styleme: Towards Intelligent Fashion Generation with Designer Style

   Di Wu^*, Zhiwang Yu^*, Nan Ma^*, Jianan Jiang, Yuetian Wang, and 3 more authors

   In Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems, 2023

   Abs DOI Code

   We introduce StyleMe, which employs Grad-CAM and AdaLIN to guide the generation of sketches, preserving the designer’s unique style in sketches. Additionally, it utilizes a two-stage generation network, combining AE and GAN, to color the generated sketches with a personalized style.