Recent advances in diffusion-based text-to-image models have simplified creating high-fidelity images, but preserving the identity (ID) of specific elements, like a personal dog, is still challenging. Object customization, using reference images and textual descriptions, is key to addressing this issue. Current object customization methods are either object-specific, requiring extensive fine-tuning, or object-agnostic, offering zero-shot customization but limited to specialized domains. The primary issue of promoting zero-shot object customization from specific domains to the general domain is to establish a large-scale general ID dataset for model pre-training, which is time-consuming and labor-intensive. In this paper, we propose a novel pipeline to construct a large dataset of general objects and build the Multi-Category ID-Consistent (MC-IDC) dataset, featuring 315k text-image samples across 10k categories. With the help of MC-IDC, we introduce Customizing Anything (CustAny), a zero-shot framework that maintains ID fidelity and supports flexible text editing for general objects. CustAny features three key components: a general ID extraction module, a dual-level ID injection module, and an ID-aware decoupling module, allowing it to customize any object from a single reference image and text prompt. Experiments demonstrate that CustAny outperforms existing methods in both general object customization and specialized domains like human customization and virtual try-on. Our contributions include a large-scale dataset, the CustAny framework and novel ID processing to advance this field. The official project page is in https://lingjiekong-fdu.github.io.

Although diffusion-based image virtual try-on has made considerable progress, emerging approaches still struggle to effectively address the issue of hand occlusion (i.e., clothing regions occluded by the hand part), leading to a notable degradation of the try-on performance. To tackle this issue widely existing in real-world scenarios, we propose VTON-HandFit, leveraging the power of hand priors to reconstruct the appearance and structure for hand occlusion cases. Firstly, we tailor a Hand-Pose Aggregation Net using the ControlNet-based structure explicitly and adaptively encoding the global hand and pose priors. Besides, to fully exploit the hand-related structure and appearance information, we propose Hand-feature Disentanglement Embedding module to disentangle the hand priors into the hand structure-parametric and visual-appearance features, and customize a masked cross attention for further decoupled feature embedding. Lastly, we customize a hand-canny constraint loss to better learn the structure edge knowledge from the hand template of model image. VTON-HandFit outperforms the baselines in qualitative and quantitative evaluations on the public dataset and our self-collected hand-occlusion Handfit-3K dataset particularly for the arbitrary hand pose occlusion cases in real-world scenarios. Our project page is at: https://vton-handfit.github.io.

Coal mining activities will result in long-term multi-level surface deformation, causing significant damage to the infrastructure and environment. Currently, the three-dimensional (3-D) mining deformation could be resolved based on single-track or multi-track Interferometric Synthetic Aperture Radar (InSAR). Before starting to monitor a study area, it is necessary to determine what sort of methodology would best meet the accuracy requirement of 3-D deformation. To accomplish this, we present an error propagation model that can clarify the relationship of error propagation between the multi-track measurements in the radar line-of-sight (LOS) or azimuth directions and the 3-D deformation in the vertical, east, and north directions. Simulation results show that the accuracy of our proposed model could reach 1.2mm, 1.3mm, and 11.6m in the vertical, east, and north directions, respectively. Accordingly, we give the precision assessment of five cases for resolving 3-D mining deformation by using our proposed model.

The combination of the LOS and azimuth measurements at ascending and descending orbits could improve the three-dimensional (3-D) deformation precision, with the accuracy dependent on the magnitude of stochastic error within InSAR measurements. Accordingly, we present the simulations by using a new 3-D deformation inversion method called CovRM-InSAR. In the case study, we applied four InSAR measurements with varying standard deviation (SD) in the LOS and azimuth directions from two tracks to map 3-D deformation. The results show that it is beneficial for improving the east and north deformation whether considering the correlation between azimuth measurements in the covariance matrix. The precision of 3-D deformation resolved by CovRM-InSAR has close accuracy in the vertical and north directions and improved by up to 88% in the east direction when compared with the least square (LS).

Three-dimensional (3-D) deformation could be resolved using multi-track Interferometric Synthetic Aperture Radar (InSAR), with the accuracy dependent on the magnitude of multi-source errors within InSAR measurements. To improve the precision of 3-D deformation, it is essential to understand the error propagation mechanism and then develop the methodology for reducing error impacts in 3-D decomposition processing. In this article, we present an error propagation model that incorporates both systematic and stochastic error propagation, which determines the error contribution of the multi-track InSAR measurements in the 3-D direction. The systematic error propagation includes generic systematic error and additional systematic errors (ASE) in the vertical and east directions caused by neglecting the north component. For stochastic error propagation, we construct the covariance matrix by considering variance and correlation from different InSAR measurements when using differential and multi-temporal InSAR techniques. Accordingly, we propose a new 3-D deformation inversion method, combining the covariance matrix and L2-norm regularization based on multi-track InSAR (CovRM-InSAR) to improve the precision of 3-D deformation with noise reduction. In the case study, we applied Sentinel-1A and ALOS-2 InSAR datasets from four tracks to map 3-D velocity in Wuhai and analyzed the time-series error propagation and 3-D uncertainty. The precision of 3-D deformation resolved by CovRM-InSAR has improved by up to 90%, 44%, and 98% in the vertical, east, and north directions, respectively. Additionally, the CovRM-InSAR has effectively reduced the stochastic errors by up to 38%, 15%, and 90% in the vertical, east, and north directions, respectively.

Model generalization to the unseen scenes is crucial to real-world applications, such as autonomous driving, which requires robust vision systems. To enhance the model generalization, domain generalization through learning the domain-invariant representation has been widely studied. However, most existing works learn the shared feature space within multi-source domains but ignore the characteristic of the feature itself (e.g., the feature sensitivity to the domain-specific style). Therefore, we propose the Domain-invariant Representation Learning (DIRL) for domain generalization which utilizes the feature sensitivity as the feature prior to guide the enhancement of the model generalization capability. The guidance reflects in two folds: 1) Feature re-calibration that introduces the Prior Guided Attention Module (PGAM) to emphasize the insensitive features and suppress the sensitive features. 2): Feature whiting that proposes the Guided Feature Whiting (GFW) to remove the feature correlations which are sensitive to the domain-specific style. We construct the domain-invariant representation which suppresses the effect of the domain-specific style on the quality and correlation of the features. As a result, our method is simple yet effective, and can enhance the robustness of various backbone networks with little computational cost. Extensive experiments over multiple domains generalizable segmentation tasks show the superiority of our approach to other methods.