National Key Laboratory of Automatic Target Recognition, College of Electronic Science and Technology, National University of Defense Technology, Changsha, China
Auto-generated profile, based on publicly available information
Yu Liu
National Key Laboratory of Automatic Target Recognition, College of Electronic Science and Technology, National University of Defense Technology, Changsha, China
It is difficult to solve Magnetic Resonance (MR) image reconstruction problems with linear combinations of total variation and ℓ1 norm regularization terms. In order to solve these compound regularization problems, we propose an efficient algorithm in this letter. The proposed algorithm adopts the Bregman iteration technique to convert the original constrained problem to a sequence of unconstrained problems, which are then solved using operator splitting and variable splitting techniques. Simulation experiments demonstrate that significant improvement of the quality of reconstructed images is achieved by the proposed algorithm when compared to previous methods.
11 September 2022 in Journal of Geophysical Research: Solid Earth
Pre-training is essential for enhancing performance across various downstream tasks, yet pre-training techniques for remote sensing image object detection remain limited. This is due to the distinct characteristics of remote sensing images, such as fixed imaging angles, complex backgrounds, and diverse target categories, which make them different from natural images. To address these challenges, this paper introduces MSP-O, an innovative two-stage mask pre-training method. Traditional single-stage pre-training frameworks aim to extract intrinsic features from large-scale datasets, improving generalization across downstream tasks. MSP-O builds upon this by adding a second-stage pre-training phase, which optimizes the model’s ability to focus on features crucial for detecting objects in downstream tasks. This additional stage allows the model to persistently acquire features beneficial for remote sensing object detection. Specifically, MSP-O incorporates a pixel-level restoration task focused on the objects of interest, guiding the model to prioritize the most critical areas for detection. This approach significantly enhances the model’s ability to discern key target features, improving both recognition precision and localization accuracy in subsequent detection tasks. Extensive experiments on the DOTA and DIOR datasets demonstrate the effectiveness of MSP-O, showing substantial improvements in detection performance compared to conventional methods.