Millimeter wave full-digital full-connected multibeam array can play a crucial role in meeting the massive data capacity demands for the future Internet of Things. In this paper, a novel digital predistortion (DPD) technique is designed to linearize this array for the scenario under constraint bandwidth. Two different linearization schemes, including beam-oriented DPD schemes and PA-oriented DPD schemes, are analyzed and compared, along with their respective applicable scenarios. Based on the characteristics of full-digital full-connected arrays, and by fully leveraging the alleviation of adjacent channel power ratio metrics requirement in FR2 band specification, the band-limited DPD concept can be effectively integrated into the full-digital full-connected multibeam array architecture, which allows for the array linearization with low system bandwidth requirements without introducing an analog filter for each PA. To demonstrate the effectiveness of the proposed technique, the simulations are analyzed for an array with 6-beam 64-chain configuration. Furthermore, experiments on a 2-beam 4-chain full-digital full-connected array are verified at the center frequency of 26 GHz with different modulated bandwidth scenarios. The experiment results indicate that the proposed method successfully achieves expected linearization performance for wideband multibeam array.

Modern deep hashing methods rely on data augmentation to fully realize their potential in the face of large-scale retrieval galleries and over-parameterized visual models. However, this work observes that mainstream label-preserving augmentation methods are unreliable for information retrieval because they lead to incomplete alignment between data and labels. This misalignment impairs metric losses in distinguishing original/augmented data during same-class clustering, compromising nearest-neighbor search efficacy. To address these issues, we propose an innovative plug-and-play data augmentation framework tailored for retrieval tasks, based on the concept ofGaining robust features by randomlyGiving up parts of the image (GG). Inspired by the ease with which visual changes induced by discard transformations can be estimated, we design two intuitive augmentation methods along with corresponding semantic shift estimators to measure the semantic changes introduced by each operation. Additionally, we optimize the metric loss based on the semantic retention scores, guiding the metric objective to properly allocate gradients for generated samples. This adjustment mitigates the adverse effects caused by incomplete alignment, optimizing the intra-class distance of both original and augmented data in the Hamming space, while ensuring the relevance and accuracy of the retrieval results. Extensive experiments conducted on six datasets demonstrate the effectiveness and robustness of our proposed framework. Code is available athttps://github.com/wuhulahu/GG.

ABSTRACT To solve efficient image segmentation in practical medical applications in resource‐constrained point‐of‐care environments, the lightweight medical image segmentation network is proposed based on grouped feature fusion and large kernel convolution, which introduces a U‐shaped, convolution‐based architecture that significantly reduces parameters and computational cost. The proposed model combines shifted large kernel convolution with grouped feature fusion technique in a lightweight and attention‐free way, which is specifically designed to fuse features to capture global context. Meanwhile, the grouped multi‐scale feature fusion module is proposed to achieve effective cross‐layer connectivity and efficient fusion of multi‐scale features by grouping deep and shallow features and subsequently applying a lightweight grouped large kernel convolution. The extensive experiments on multiple datasets verify that our model outperforms current popular models in image segmentation with lower parameter quantity and computational cost, and achieves industry‐leading performance with low resource consumption.

ABSTRACT Fundus photography provides imaging techniques for the diagnosis of retinal diseases. The diagnostic accuracy, however, heavily relies on the clarity of subtle lesions, which can be significantly affected by image resolution. Achieving a balance between reconstruction quality, model complexity, and training efficiency remains a key challenge, particularly under limited data conditions. To address these issues, a lightweight end‐to‐end model LiteZSSR is proposed for super‐resolution reconstruction of fundus images, incorporating a residual information distillation module to extract multi‐scale features within a shallow network architecture, effectively retaining both local and global contextual information. In addition, a multi‐feature fusion group composed of multiple large kernel attention blocks is designed to strengthen feature representation while minimizing redundancy and computational overhead. Unsupervised training based on internal image learning is adopted to eliminate dependence on large‐scale datasets and to suppress artifacts commonly produced by CNN‐based SRR methods. Extensive experiments on publicly available fundus image datasets, including DRIVE, STARE, and CHASEDB1, demonstrate that LiteZSSR outperforms existing state‐of‐the‐art methods in terms of PSNR and SSIM, while significantly reducing model parameters. These results highlight its potential for practical deployment in clinical fundus image enhancement tasks.

Recently, ShiftViT and its variants have attracted much attention for their simple and efficient shift operation, showing excellent efficacy in several tasks on natural images, surpassing Swin Transformer. However, considering the complexity of 3D multimodal images, which have higher dimensions than natural images, and the relative stability of the human tissue structure in medical images, the applicability of shift operation on 3D multimodal medical data has yet to be determined. This paper demonstrates that ShiftViT has enormous potential in 3D multimodal medical image analysis. Using 3D medical image segmentation as a representative downstream task, we investigate how shift operation can improve model performance. First, applying ShiftViT to 3D multimodal medical images not only effectively extracts global information but also significantly enhances the model's performance. Second, as a plug-and-play strategy, the shift operation can be integrated with other modules without adding additional computational burden, proving its flexibility in the overall system. Finally, we further investigate the generalizability of the shift operation by introducing a cascaded attention module, which provides useful insights to improve the generalizability of 3D medical image segmentation models. Through this study, we extend the application scope of ShiftViT and bring new exploration directions to the field of 3D multimodal medical image analysis. Our research results prove the feasibility of applying ShiftViT in 3D multimodal medical images and provide an effective and scalable model, which is expected to further promote the development of medical image processing technology.