To minimize knowledge interference during fusion, this work presents a gradient-based orthogonal refreshing strategy that projects gradient updates of new domains onto the orthogonal complement of the fused historical subspace, supporting continual adaptation without forgetting.
Abstract
Domain-incremental object detection (DIOD) requires models to continually adapt to new domains while preserving prior knowledge. Recently, parameter-efficient fine-tuning offers a promising avenue, wherein a pre-trained model is frozen and a small number of learnable parameters are injected for downstream tasks. However, these methods risk overwriting critical past knowledge, triggering inter-domain interference and performance degradation. To address this challenge, we propose Orthogonal Knowledge Refreshing (OKR), a simple yet effective framework for DIOD. OKR incrementally constructs independent domain-specific subspaces via dedicated low-rank branches for each domain, which are seamlessly fused for a holistic decision, enabling conflict-free capacity expansion without domain selection during inference. To minimize knowledge interference during fusion, we present a gradient-based orthogonal refreshing strategy that projects gradient updates of new domains onto the orthogonal complement of the fused historical subspace, supporting continual adaptation without forgetting. Moreover, to mitigate semantic fragmentation across domains, we enforce topology-aware consistency, aligning the semantic structures of old and new domains. Extensive experiments validate the superiority of OKR, outperforming the best exemplar-free method by significant margins of +5.6% and +6.5% mAP on the Pascal VOC and BDD100K series, respectively.
DAER maintains domain-isolated lightweight experts while keeping the pretrained Video-MLLM backbone frozen, thereby decoupling domain-specific adaptation from the general multimodal knowledge of the pretrained model, and introduces adaptive domain merging to improve parameter scalability and a two-stage optimization strategy to stabilize expert specialization during continual learning.
Rui Cheng, Meixing Shi, Yuxiang Cai et al.· arXiv.org· 0 citations
Vision-language object detectors (VLODs) pretrained on large image-text corpora exhibit strong zero-shot detection abilities, yet their performance often degrades on specialized downstream tasks that arrive sequentially. Hence, we study Incremental Vision-Language Object Detection (IVLOD), which requires continually adapting a VLOD across tasks with domain and/or class shifts while mitigating catastrophic forgetting of prior tasks and preserving its zero-shot generalization. To address this problem, we propose a novel method, named Sparse Frequency Transform (SFT), that minimizes inter-task interference via a frequency-domain design. Building on a theoretical link between forgetting and Frobenius inner product (FIP), SFT enforces sparsity-independence on spectral supports, yielding zero inter-task FIP by construction and achieving reduced forgetting with low impact on the pretrained model. Spectral supports are selected via an online gradient-magnitude scoring rule with masking of previously used supports, then converted via inverse discrete cosine transform (IDCT) to dense time-domain updates that preserve FIP constraints and improve optimization efficiency. Extensive experiments under both full-shot and few-shot IVLOD settings demonstrate that our SFT is capable of consistently learning new tasks while preserving the zero-shot generalization capabilities of the pretrained model.
Xiang Song, Zi-Ye Yang, Yuhang He et al.· IEEE Transactions on Image P...· 0 citations
Vision foundation models (VFMs) offer strong generalization capabilities for domain-adaptive object detection (DAOD). However, existing VFM-based methods overlook the spatial-scale discrepancy between teacher and student feature maps, resulting in semantic incompatibility that weakens both feature alignment and pseudo-label learning. Moreover, domain shift can cause source-trained VFM teachers to miss target-domain objects, limiting the quality of their pseudo-labels. To address these issues, we propose the Semantic Localization-Enhanced Teacher (SLE-T), a semantically compatible knowledge-distillation framework built around a lightweight SLE Adapter for DINOv2. SLE Adapter injects pretrained local-texture priors into DINOv2 to improve cross-domain recognition and reformulates its features into dense representations that are spatially and semantically compatible with the student detector. SLE-T transfers the resulting teacher knowledge through either pseudo-label learning or feature alignment. We instantiate SLE-T with DINOv2-B and DINOv2-L (the ViT-B and ViT-L variants) and compare them with the larger DINOv2-G teacher. Extensive experiments on three DAOD benchmarks demonstrate that our method achieves state-of-the-art performance, and ablation studies confirm the importance of teacher-student semantic compatibility. Notably, SLE-T with DINOv2-B produces competitive or superior pseudo-labels using approximately one-quarter of the training time of DINOv2-G and substantially less GPU memory, demonstrating efficient VFM knowledge transfer under limited computational resources.
Qifeng Zhang, Ting Xiang, Ze-Yu Bai et al.· 0 citations
Source-free domain adaptation (SFDA) aims to adapt pre-trained source models to new target domains without requiring access to any source domain data, thereby addressing privacy and efficiency concerns. Existing SFDA methods for object detection primarily follow a teacher–student self-training paradigm; however, their performance is often limited by noisy pseudo-labels. To address this issue, this paper proposes an SFDA object detection framework for the YOLO family of single-stage detectors. First, a weak–strong pseudo-label consistency filtering strategy is designed to remove unreliable pseudo-labels by exploiting the prediction consistency across different augmented views. Second, a multiscale object-level contrastive learning mechanism is introduced to extract object-level features at multiple feature scales, thereby enhancing the consistency and discriminability of object representations across different views and scales through supervised contrastive constraints. Experimental results show that the proposed method consistently outperforms the baseline on multiple cross-domain detection tasks, demonstrating its effectiveness and good generalization ability under the source-free setting.
As the de facto standard for knowledge-intensive tasks, Retrieval-Augmented Generation (RAG) has significantly enhanced the reliability of Large Language Models by incorporating external non-parametric knowledge. However, adapting general-purpose retrievers to specific vertical domains often triggers catastrophic forgetting, severely degrading performance on open-domain queries. Additionally, existing mitigation strategies, such as linear model fusion, are mathematically constrained within a linear geometric manifold, limiting their ability to effectively rectify complex non-linear semantic drifts caused by domain shifts. To address these limitations, we present an effective approach, SuRe-EM (Subspace-routing & Residual-corrected Expert Model), designed to resolve the issues of domain specialization and Cross-Domain generalization in domain-adaptive retrieval. SuRe-EM enhances Cross-Domain representation by integrating fine-grained subspace routing with non-linear residual correction. Specifically, SuRe-EM first employs subspace routing to dynamically decouple high-dimensional features for maximizing domain specialization, followed by a residual module that generates non-linear semantic compensations. We validate our model on a vertical domain dataset (AHD) and general domain datasets (CMRC, SQuAD). Quantitative results demonstrate the effectiveness of SuRe-EM, which maintains strong In-Domain precision while improving Recall@10 by up to 4.46 points over state-of-the-art linear fusion baselines in Cross-Domain scenarios. Furthermore, comprehensive ablation studies validate the non-redundant synergy of the key design elements within SuRe-EM.
Xi-Fang Liu· Poster Volume 0007 The 2026...· 0 citations
Efficient Unsupervised Domain Adaptation (EUDA) is proposed, a parameter-efficient framework that leverages a frozen DINOv2 backbone as a feature extractor and updates only a lightweight bottleneck and classification head to promote both discriminative learning and cross-domain alignment.
Ali Abedi, Q. M. J. Wu, Ning Zhang et al.· International Journal of Mac...· 9 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
