Domain heterogeneity in federated learning presents significant challenges, particularly in complex tasks like semantic segmentation, where pixel-level accuracy is crucial. Existing approaches primarily focus on extracting domain-specific features, but they often overlook the importance of class-level feature invariance. In this paper, we propose FedSS, a novel federated semantic segmentation framework designed to address domain heterogeneity. FedSS introduces a Fine-grained Domain-agnostic Feature Extraction (DFE) module that standardizes feature maps using category-level statistics to ensure domain-agnostic feature extraction. Additionally, the Adaptive Ordering Based Feature Decorrelation (AFD) module enhances the model’s ability to distinguish between different semantic categories by decorrelating feature channels. To further tackle domain discrepancies, we present a Domain Fairness-aware Aggregation Strategy (DFA) that dynamically adjusts aggregation weights based on local domain variations. Our approach improves the robustness and generalization of the semantic segmentation model across diverse domains, ensuring more accurate and reliable pixel-level predictions. Experimental results demonstrate the effectiveness of FedSS in addressing domain heterogeneity and enhancing segmentation performance in federated learning settings. The source code is released at https://github.com/vibratingwings/FedSemanSeg

Federated Class Incremental Learning (FCIL) has emerged as a new paradigm due to its applicability in real-world scenarios. In FCIL, clients continuously generate new data with unseen class labels and do not share local data due to privacy restrictions, and each client’s class distribution evolves dynamically and independently. However, existing work still faces two significant challenges. Firstly, current methods lack a better balance between maintaining sound anti-forgetting effects over old data (stability) and ensuring good adaptability for new tasks (plasticity). Secondly, some FCIL methods overlook that the incremental data will also have a non-identical label distribution, leading to poor performance. This paper proposes CGoFed, which includes relax-constrained gradient update and cross-task gradient regularization modules. The relax-constrained gradient update prevents forgetting the knowledge about old data while quickly adapting to the new data by constraining the gradient update direction to a gradient space that minimizes interference with historical tasks. The cross-task gradient regularization also finds applicable historical models from other clients and trains a personalized global model to address the non-identical label distribution problem. The results demonstrate that the CGoFed performs well in alleviating catastrophic forgetting and improves model performance by 8% -23% compared with the SOTA comparison method.

Human activity recognition (HAR) is now a powerful in-home fitness assistive technology. This article presents a wearables-based multisensor system (WMS), which not only supports conventional functionalities, such as motion evaluation based on multidimensional information about the user’s body (movement speed, angle, muscle states, etc.), but also provides advanced services, including assessing training fatigue and providing real time, elastic, and professional training advice. The proposed WMS is experimentally validated by yielding 90.11% accuracy of motion evaluation with 36% and 23% improvement of fitness effect on bicep girth and muscular endurance, indicating its feasibility to prompt the development of HAR in the in-home fitness training domain.