Remote sensing change detection(RSCD) aims to identify the regions of interest that have changed between dual-temporal images. However, most deep models predict CD results by extracting multi-scale hybrid features, which could easily lead to ambiguous semantic boundaries; in addition, the existing feature acquisition tends to lack consideration of capturing their diversity usually causing poor model generalization. Thus, this paper decomposes the mixed features into change and invariant features jointly with stochastic distribution sampling and convolution thus accomplishing robust RSCD based on decoupled representations. In the training stage, the posterior distribution of the uncoupled features is first learned through label calibration to train the prior distribution generator; then, robust feature decoupling is implemented combining the convolutional feature separator with re-parameterized sampling over the decoupled posteriori distribution, and further aggregating the decoupled features through prototype learning; finally, the exceed-expectation loss regularizer is proposed to push or pull these positive and negative sample features to a more distant end, thereby increasing the inter-class distance by boosting the predicted expectation. In the testing stage, the robust RSCD based on decoupled representation is accomplished through the feature separator, decoupled prior distribution random sampling, and CD head without posterior distribution support. Experiments prove that DSFDcd has achieved remarkable results in terms of qualitative and quantitative metrics. Our codes will be available at https://github.com/iceking111/DSFDcd.

Despite the remarkable success of foundation models, their task-specific fine-tuning paradigm makes them inconsistent with the goal of general perception modeling. The key to eliminating this inconsistency is to use generalist models for general task modeling. However, existing attempts at generalist models are inadequate in both versatility and performance. In this paper, we propose Uni-Perceiver v2, which is the first generalist model capable of handling major large-scale vision and vision-language tasks with competitive performance. Specifically, images are encoded as general region proposals, while texts are encoded via a Transformer-based language model. The encoded representations are transformed by a task-agnostic decoder. Different tasks are formulated as a unified maximum likelihood estimation problem. We further propose an effective optimization technique named Task-Balanced Gradient Normalization to ensure stable multi-task learning with an unmixed sampling strategy, which is helpful for tasks requiring large batch-size training. After being jointly trained on various tasks, Uni-Perceiver v2 is capable of directly handling downstream tasks without any task-specific adaptation. Results show that Uni-Perceiver v2 outperforms all existing generalist models in both versatility and performance. Meanwhile, compared with the commonly-recognized strong baselines that require tasks-specific fine-tuning, Uni-Perceiver v2 achieves competitive performance on a broad range of vision and vision-language tasks.

Knowledge distillation has been applied to image classification successfully. However, object detection is much more sophisticated and most knowledge distillation methods have failed on it. In this paper, we point out that in object detection, the features of the teacher and student vary greatly in different areas, especially in the foreground and background. If we distill them equally, the uneven differences between feature maps will negatively affect the distillation. Thus, we propose Focal and Global Distillation (FGD). Focal distillation separates the foreground and background, forcing the student to focus on the teacher's critical pixels and channels. Global distillation rebuilds the relation between different pixels and transfers it from teachers to students, compensating for missing global information in focal distillation. As our method only needs to calculate the loss on the feature map, FGD can be applied to various detectors. We experiment on various detectors with different backbones and the results show that the student detector achieves excellent mAP improvement. For example, ResNet-50 based RetinaNet, Faster RCNN, RepPoints and Mask RCNN with our distillation method achieve 40.7%, 42.0%, 42.0% and 42.1% mAP on COCO2017, which are 3.3, 3.6, 3.4 and 2.9 higher than the baseline, respectively. Our codes are available at https://github.com/yzd-v/FGD.

Source positioning is a key issue in a range of applications such as sensing, monitoring and tracking, etc. In this paper, we address the source localization problem with unknown clock skew and outliers in the TOA measurements. Considering the measurement outliers, we present a robust localization formulation by introducing Huber loss. Furthermore, we develop a lightweight iterative localization algorithm using majorization-minimization (MM) method with guaranteed convergence. In addition, we propose a novel semidefinite-relaxation-based algorithm for the proposed robust localization formulation. Simulations demonstrate that our proposed MM-based localization algorithms could achieve better performance than SDR-based localization algorithms in terms of both localization accuracy and computational time, for both cases with or without outliers.