News
2022-03: Invitated as a reviewer for NeurIPS 2022.
2022-01: One paper is accepted by ICLR 2022.
2021-09: One paper is accepted by NeurIPS 2021.
2021-07: One paper is accepted by ICCV 2021 and one paper is accepted by TIP.
|
|
Bridging the Source-to-target Gap for Cross-domain Person Re-Identification with Intermediate Domains
Yongxing Dai,
Yifan Sun,
Jun Liu,
Zekun Tong,
Yi Yang,
Ling-Yu Duan
Arxiv preprint, 2022
paper
/
code
/
bibtex
IDM++ is a journal extension of IDM (ICCV 2021 Oral). We further propose a Mirrors Generation Module (MGM) to reinforce IDM into IDM++. IDM++ provides a unified framework
for two popular cross-domain re-ID scenarios (i.e., UDA and DG). Twelve UDA re-ID benchmarks and two DG re-ID protocols validate that IDM++ brings general improvement and
sets new state of the art for both scenarios.
|
|
Uncertainty Modeling for Out-of-Distribution Generalization
Xiaotong Li,
Yongxing Dai,
Yixiao Ge,
Jun Liu,
Ying Shan,
Ling-Yu Duan
ICLR, 2022
paper
/
code
/
bibtex
We improve the network generalization ability by modeling the uncertainty
of domain shifts with synthesized feature statistics during training. Our method shows its superiority on a wide range of out-of-distribution (OOD) vision tasks including image classification, semantic segmentation, and instance retrieval.
|
|
Digraph Contrastive Learning
Zekun Tong,
Yuxuan Liang,
Henghui Ding,
Yongxing Dai,
Xinke Li,
Changhu Wang,
NeurIPS, 2021
paper
/
code
/
bibtex
We design a digraph data augmentation method called Laplacian perturbation and theoretically
analyze how it provides contrastive information without changing the digraph structure.
|
|
IDM: An Intermediate Domain Module for Domain Adaptive Person Re-ID
Yongxing Dai,
Jun Liu,
Yifan Sun,
Zekun Tong,
Chi Zhang,
Ling-Yu Duan
ICCV, 2021   (Oral Presentation)
paper
/
arXiv
/
code
/
bibtex
IDM is a plug-and-play module to generate intermediate domains' representations,
which can bridge source and target domains in unsupervised domain adaptive person re-ID.
|
|
Dual-Refinement: Joint Label and Feature Refinement for Unsupervised Domain Adaptive Person Re-Identification
Yongxing Dai,
Jun Liu,
Yan Bai,
Zekun Tong,
Ling-Yu Duan
IEEE Transactions on Image Processing (TIP), 2021  
paper
/
arXiv
/
code
/
bibtex
Dual-Refinement can jointly refine pseudo labels at the off-line clustering phase and features at the on-line training phase, to alternatively boost the label purity and feature discriminability in the target domain for
more reliable re-ID.
|
|
Generalizable Person Re-identification with Relevance-aware Mixture of Experts
Yongxing Dai,
Xiaotong Li,
Jun Liu,
Zekun Tong,
Ling-Yu Duan
CVPR, 2021  
paper
/
arXiv
/
bibtex
We propose a novel method called the relevance-aware mixture of experts (RaMoE),
using an effective voting-based mixture mechanism to dynamically leverage source domains’
diverse characteristics to improve the model’s generalization
|
|
Disentangled Feature Learning Network for Vehicle Re-Identification
Yan Bai,
Yihang Lou,
Yongxing Dai,
Jun Liu,
Ziqian Chen,
Ling-Yu Duan
IJCAI, 2020  
paper
/
bibtex
We propose a Disentangled Feature Learning Network (DFLNet) to learn orientation specific and common features
concurrently. Moreover, to effectively use these two types of features for ReID, we further design a feature metric
alignment scheme to ensure the consistency of the metric scales.
|
Academic Services
Journal Reviewer:TIP, TMM, Neurocomputing
Conference Reviewer:NeurIPS 2022, ICLR 2022
|
|