Image restoration aims to reconstruct a sharp image from its degraded counterpart, which plays an important role in many fields. Recently, Transformer models have achieved promising performance on various tasks. However, their quadratic complexity remains intractable issue for practical applications. The aim of this study is develop efficient and effective framework restoration. Inspired by the fact that different regions corrupted always undergo degradations degrees, we propose focus more...

Image restoration aims to reconstruct the latent sharp image from its corrupted counterpart. Besides dealing with this long-standing task in spatial domain, a few approaches seek solutions frequency domain by considering large discrepancy between spectra of sharp/degraded pairs. However, these algorithms commonly utilize transformation tools, e.g., wavelet transform, split features into several parts, which is not flexible enough select most informative component recover. In paper, we...

Image restoration aims to reconstruct a high-quality image from degraded low-quality observation. Recently, Transformer models have achieved promising performance on tasks due their powerful ability model long-range dependencies. However, the quadratically growing complexity with respect input size makes them inapplicable practical applications. In this paper, we develop an efficient convolutional network for by enhancing multi-scale representation learning. To end, propose omni-kernel...

Image restoration aims to reconstruct a high-quality image from its corrupted version, playing essential roles in many scenarios.Recent years have witnessed paradigm shift convolutional neural networks (CNNs) Transformerbased models due their powerful ability model long-range pixel interactions.In this paper, we explore the potential of CNNs for and show that proposed simple network architecture, termed ConvIR, can perform on par with or better than Transformer counterparts.By re-examing...

As a long-standing and challenging task, image deblurring aims to reconstruct the latent sharp from its degraded counterpart. In this study, bridge gaps between degraded/sharp pairs in spatial frequency domains simultaneously, we develop dual-domain attention mechanism for deblurring. Self-attention is widely used vision tasks, however, due quadratic complexity, it not applicable with high-resolution images. To alleviate issue, propose novel module by implementing self-attention style of...

Several observational studies have proposed a potential link between gut microbiota and the onset progression of sepsis. Nevertheless, causality sepsis remains debatable warrants more comprehensive exploration.

As a long-standing task, image restoration aims to recover the latent sharp from its degraded counterpart. In recent years, owing strong ability of self-attention in capturing long-range dependencies, Transformer based methods have achieved promising performance on multifarious tasks. However, canonical leads quadratic complexity with respect input size, hindering further applications restoration. this paper, we propose Strip Attention Network (SANet) for integrate information more efficient...

Image restoration aims to recover a clean image from various degradations, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">e.g.,</i> haze, snow, and blur, playing an important role in robot vision, autonomous vehicles, medical imaging. Recently, the use of Transformer has witnessed significant improvement multifarious tasks. However, despite few remedies reduce quadratic complexity self-attention, these approaches are still impractical for...

Autonomous driving has rapidly developed and shown promising performance with recent advances in hardware deep learning methods. High-quality datasets are fundamental for developing reliable autonomous algorithms. Previous dataset surveys tried to review the but either focused on a limited number or lacked detailed investigation of characters datasets. To this end, we present an exhaustive study over 200 from multiple perspectives, including sensor modalities, data size, tasks, contextual...

Vision systems are the core element in industrial systems, such as intelligent transportation and inspection robots. However, undesired degradations caused by bad weather or low-end devices reduce visibility of images. Image restoration aims to reconstruct a sharp image from degraded counterpart plays an important role systems. Recent transformer-based architectures leverage self-attention unit convolutions model long-range dependencies local connectivity, respectively, achieving promising...

With the rapid development of virtual reality, omnidirectional images (ODIs) have attracted much attention from both industrial community and academia. However, due to storage transmission limitations, resolution current ODIs is often insufficient provide an immersive reality experience. Previous approaches address this issue using conventional 2D super-resolution techniques on equirectangular projection without exploiting unique geometric properties ODIs. In particular, (ERP) provides a...

Image restoration involves recovering a high-quality image from its corrupted counterpart. This paper presents an effective and efficient framework for restoration, termed CSNet, based on ``channel + spatial" hybrid frequency modulation. Different feature channels include different degradation patterns degrees, however, most current networks ignore the importance of channel interactions. To alleviate this issue, we propose frequency-based modulation module to facilitate interactions through...

In the image acquisition process, various forms of degradation, including noise, haze, and rain, are frequently introduced. These degradations typically arise from inherent limitations cameras or unfavorable ambient conditions. To recover clean images degraded versions, numerous specialized restoration methods have been developed, each targeting a specific type degradation. Recently, all-in-one algorithms garnered significant attention by addressing different types within single model...

This paper proposes a novel single shot network for object detection. The proposed network, termed IDNet, explores the strategies of feature fusion to alleviate scale variation problem in IDNet mainly consists two modules: an indirect module (IF) and direct (DF). IF shares long-range dependencies within pyramidal layers based on these information, learns emphasize informative regions suppress less useful ones each layer. DF is strategy modified lateral connection inspired by pyramid networks...

Facing the scale variation challenge in topic of object detection, this papers we design feature fusion methods to improve representation ability features. The proposed network, Feature Fusion Network (FFNet), contains mainly two parts: Relationship Module (RFM) and Numerical (NFM). long dependencies information is shared within pyramidal features RFM. This helps each emphasize informative regions reduce influence useless regions. NFM introduces averaging operation generate weights retain...