Inpainting deteriorated regions in digital Dunhuang murals is important for Dunhuang mural content preservation. Algorithms of mural image inpainting help simplify the digital restoration process of the deteriorated murals. Most of the existing algorithms can restore plausible content for homogeneous missing regions in Dunhuang mural images. However, they often fail to fill accurate color in missing regions that contain complex structures, which is mainly due to the neglect of color relevance between positions in the missing structural region and the non-missing color regions. In this article, we propose a deep learning–based, structure-guided inpainting method for the Dunhuang mural image, which utilizes relevant color information in deep features to improve the color inpainting quality for structural regions. Specifically, we design a structure-guided feature refinement module, which explicitly leverages color relevance implied in structure information to select relevant features for refining features in the missing region. In addition, we propose a multi-step scheme for feature refinement to better propagate non-missing region feature information to the missing region. We conduct experiments on Dunhuang660 and Dunhuang No. 7 Grotto datasets. The results demonstrate that our proposed method can achieve improved color inpainting quality for missing structural regions in Dunhuang mural images.

This paper addresses two vital challenges in Unsupervised Domain Adaptation (UDA) with a focus on harnessing the power of Vision-Language Pre-training (VLP) models. Firstly, UDA has primarily relied on ImageNet pre-trained models. However, the potential of VLP models in UDA remains largely unexplored. The rich representation of VLP models holds significant promise for enhancing UDA tasks. To address this, we propose a novel method called Cross-Modal Knowledge Distillation (CMKD), leveraging VLP models as teacher models to guide the learning process in the target domain, resulting in state-of-the-art performance. Secondly, current UDA paradigms involve training separate models for each task, leading to significant storage overhead and impractical model deployment as the number of transfer tasks grows. To overcome this challenge, we introduce Residual Sparse Training (RST) exploiting the benefits conferred by VLP’s extensive pre-training, a technique that requires minimal adjustment (approximately 0.1%~0.5%) of VLP model parameters to achieve performance comparable to fine-tuning. Combining CMKD and RST, we present a comprehensive solution that effectively leverages VLP models for UDA tasks while reducing storage overhead for model deployment. Furthermore, CMKD can serve as a baseline in conjunction with other methods like FixMatch, enhancing the performance of UDA. Our proposed method outperforms existing techniques on standard benchmarks. Our code will be available at: https://github.com/Wenlve-Zhou/VLP-UDA.

Pedestrians in videos are usually in a moving state, resulting in serious spatial misalignment like scale variations and pose changes, which makes the video-based person re-identification problem more challenging. To address the above issue, in this paper, we propose a Frame-Guided Region-Aligned model (FGRA) for discriminative representation learning in two steps in an end-to-end manner. Firstly, based on a frame-guided feature learning strategy and a non-parametric alignment module, a novel alignment mechanism is proposed to extract well-aligned region features. Secondly, in order to form a sequence representation, an effective feature aggregation strategy that utilizes temporal alignment score and spatial attention is adopted to fuse region features in the temporal and spatial dimensions, respectively. Experiments are conducted on benchmark datasets to demonstrate the effectiveness of the proposed method to solve the misalignment problem and the superiority of the proposed method to the existing video-based person re-identification methods.

Domain adaptation is still a challenging task due to the fact that the distribution discrepancy between source domain and target domain weakens the transfer ability. Intuitively, it is crucial to discover a more discriminative feature representation across domains. However, previous methods do not take the target discriminative information into account since (most) target data are unlabelled. Here, the authors propose a transfer metric learning method which decreases intra‐class distance and increases inter‐class distance simultaneously even in the case of target data are unlabelled. The shared features are more discriminative, hence the model could be more robust for target data. Specially, the global optimal solution can be obtained by solving a generalised eigen‐decomposition problem. Extensive experiments on image datasets demonstrate that compared to several state‐of‐the‐art methods, authors’ method achieves significant improvement of 9.0% in average classification accuracy.