17 May 2023 in IEEE Transactions on Circuits and Systems for Video Technology
Federated domain adaptation (FDA) is an effective method for performing over networks from distributed parties, which well improve data privacy and portability in unsupervised multi-source domain adaptation (UMDA) tasks. Despite the impressive gains achieved, two common limitations exist in current FDA works. First, most prior studies require access to the model parameters or gradient details of each source party. However, the raw source data can be reconstructed from the model gradients or parameters, which may leak individual information. Second, these works assume that different parties share an identical network architecture, which is impractical and not desirable for low- or high-resource target users. To address these issues, in this work, we point out a more practical UMDA setting, called Federated Multi-source Domain Adaptation on Black-box Models (B 2 FDA), where all data are stored locally and only the input-output interface of the source model is available. To tackle B 2 FDA, we propose an effective method, termed Co 2 -Learning with Multi-Domain Attention (Co-MDA). Experiments on multiple benchmark datasets demonstrate the efficacy of our proposed method. Notably, Co-MDA performs comparably with traditional UMDA methods where the source data or the trained model are fully available.
24 March 2024 in Proceedings of the AAAI Conference on Artificial Intelligence
Conventional Federated Domain Adaptation (FDA) approaches usually demand an abundance of assumptions, which makes them significantly less feasible for real-world situations and introduces security hazards. This paper relaxes the assumptions from previous FDAs and studies a more practical scenario named Universal Federated Domain Adaptation (UFDA). It only requires the black-box model and the label set information of each source domain, while the label sets of different source domains could be inconsistent, and the target-domain label set is totally blind. Towards a more effective solution for our newly proposed UFDA scenario, we propose a corresponding methodology called Hot-Learning with Contrastive Label Disambiguation (HCLD). It particularly tackles UFDA's domain shifts and category gaps problems by using one-hot outputs from the black-box models of various source domains. Moreover, to better distinguish the shared and unknown classes, we further present a cluster-level strategy named Mutual-Voting Decision (MVD) to extract robust consensus knowledge across peer classes from both source and target domains. Extensive experiments on three benchmark datasets demonstrate that our method achieves comparable performance for our UFDA scenario with much fewer assumptions, compared to previous methodologies with comprehensive additional assumptions.
29 June 2024 in Knowledge-Based Systems
12 August 2026 in SSRN Electronic Journal
3D Gaussian Splatting (3DGS) has recently emerged as a powerful 3D representation, providing rich Gaussian attributes that benefit downstream perception tasks (e.g., classification). However, 3DGS-based models often face inevitable domain shifts in real-world deployment, making unsupervised domain adaptation (UDA) on 3DGS—termed Gaussian Domain Adaptation (GDA)—essential. Yet, exploring the GDA scenario presents two main challenges: (i) There are no standardized evaluation benchmarks explicitly designed for GDA. (ii) 3DGS introduces heterogeneous attributes beyond coordinates with domain shifts that vary in magnitude and statistics; directly feeding all attributes into a conventional 3D UDA pipeline may therefore amplify attribute-level gaps and cause negative transfer. To address the first challenge, we developed GDA-10, a benchmark consisting of three Gaussian splatted datasets (i.e., ModelSplat-10, ShapeSplat-10, and MVImgSplat-10), each with distinct characteristics for GDA. To address the second challenge, we propose GDA-Net, a dedicated UDA Network for 3D Gaussian Splats, built upon a decouple-then-fuse strategy. At its core is an attribute-guided Mixture-of-Experts strategy (GDA-MoE), which dynamically disentangles and fuses heterogeneous 3DGS attributes to selectively emphasize the reliable, domain-invariant information. To focus on learning domain-invariant features from both the source and target domains, we then implement progressive feature and classifier adaptation based on the refined features of GDA-MoE. Extensive experiments on both single-label and Multi-Label GDA-10 demonstrate the superiority of our approach over state-of-the-art 3D-based UDA methods in terms of both classification accuracy and Macro-F1 score.