Bin Hu

Transformer architectures have demonstrated remarkable efficacy in image restoration; however, their quadratic self‐attention complexity imposes severe computational constraints for deployment on resource‐limited platforms. We investigate an alternative pathway within the transformer family itself: replacing quadratic attention with linear attention mechanisms while preserving the architectural strengths of spatial‐wise and channel‐wise feature modelling. We introduce Mixed Linear Transformer (MLT) blocks, each integrating multi‐head spatial‐wise linear self‐attention and multi‐head channel‐wise linear self‐attention, to construct a lightweight student network. To compensate for the representational capacity reduction inherent to linearization, we propose a Dual Feature‐based Knowledge Distillation (DFKD) framework that operates in an offline manner: a high‐capacity transformer‐based teacher is first pre‐trained via supervised learning and subsequently frozen, while the student learns to replicate both spatial‐wise and channel‐wise attention patterns through dedicated feature alignment losses. Extensive evaluation across multiple image restoration benchmarks demonstrates that the distilled MLT‐based model, operating with merely 1.45 M parameters and 12.8 GFLOPs, achieves competitive or superior performance relative to state‐of‐the‐art approaches.

Benefiting from the powerful nonlinear fitting ability of neural networks, deep learning‐based methods have gradually emerged as the dominate solutions for single image dehazing. However, supervised learning–based methods require paired image samples for training. To address this, an unsupervised Vision Mamba with Contrastive Regularization network (VMCR) is proposed. The network is designed based on the DisentGAN framework, and its main module is the Vision Mamba. This module performs very competitively compared to transformers, while maintaining linear time complexity and constant memory complexity with respect to the input size. Furthermore, a contrastive regularization method based on contrastive learning is proposed to enhance the reconstruction capabilities of the network and achieve superior dehazing results. Our VMCR‐Net outperforms state‐of‐the‐art unsupervised image dehazing methods, as evidenced by experimental results on several benchmarks. This research successfully proposes an enhanced unsupervised image dehazing approach, overcoming the limitations of existing methods and achieving superior dehazing performance.

Single image deraining aims at generating a clear image from its raining counterpart, and the convolutional neural network based methods have achieved great success. We propose an effective long-range attention based network for single image deraining. Existing attention module relies on current feature as input and generates attention to current feature, so we adopt the information-growth attention to enhance the current feature by using former features. Then a multi scale information-growth self-attention is proposed to extract the long-range representations, and it divides features into multiscales in attention generation process, and its lager receptive field provides more information and more accurate attention with less computation cost. The experiments on several datasets demonstrate that our method achieve better results than the recent state-of-art methods.

Addressing issues such as high parameter count and low detection accuracy in weed detection algorithms, this paper presents a study on a weed detection alg

Benefiting from the powerful nonlinear fitting ability of neural networks, deep learning-based methods have gradually emerged as the dominate solutions for single image dehazing. However, supervised learning–based methods require paired image samples for training. To address this, an unsupervised Vision Mamba with Contrastive Regularization network (VMCR) is proposed. The network is designed based on the DisentGAN framework, and its main module is the Vision Mamba. This module performs very competitively compared to transformers, while maintaining linear time complexity and constant memory complexity with respect to the input size. Furthermore, a contrastive regularization method based on contrastive learning is proposed to enhance the reconstruction capabilities of the network and achieve superior dehazing results. Our VMCR-Net outperforms state-of-the-art unsupervised image dehazing methods, as evidenced by experimental results on several benchmarks. This research successfully proposes an enhanced unsupervised image dehazing approach, overcoming the limitations of existing methods and achieving superior dehazing performance.

In the digital age, multimodal archival data is experiencing explosive growth, and how to efficiently and accurately retrieve information from it has become a key challenge. Traditional retrieval methods struggle to effectively handle multi-source heterogeneous multimodal data, leading to poor retrieval accuracy and efficiency. To address this issue, this paper proposes the MDKG-RL model, which organically integrates knowledge graph reasoning, deep reinforcement learning dynamic optimization, and multimodal Transformer architecture to achieve deep semantic understanding of multimodal data and intelligent optimization of retrieval strategies. The experiments, based on the ICDAR 2023 and AIDA Corpus datasets, show that MDKG-RL achieves a mean reciprocal rank (MRR) of 0.85, a normalized discounted cumulative gain (NDCG) of 0.88, and an entity linking accuracy of 92.4%. Compared to the baseline model, MRR improves by 13.3%, NDCG increases by 12.8%, and response time is reduced by 38.2%, significantly outperforming other comparison models. Ablation experiments also confirm the indispensability of each module. Visual analysis further demonstrates the model’s clear advantages in retrieval accuracy and efficiency, though error analysis reveals its shortcomings in handling long-tail entities and cross-modal ambiguity. The MDKG-RL model provides an innovative and effective solution for multimodal archival retrieval, not only improving retrieval performance but also laying the foundation for future research. In the future, model performance and generalization capabilities can be further enhanced by expanding data, optimizing strategies, and extending application scenarios, thereby promoting the development and application of multimodal retrieval technology in the fields of information management and knowledge discovery.

Power Line Autonomous Inspection (PLAI) plays a crucial role in the construction of smart grids due to its great advantages of low cost, high efficiency, and safe operation. PLAI is completed by accurately detecting the electrical components and defects in the aerial images captured by Unmanned Aerial Vehicles (UAVs). However, the visible quality of aerial images is inevitably degraded by adverse weather like haze, rain, or snow, which are found to drastically decrease the detection accuracy in our research. To circumvent this problem, we propose a new task of Power Line Aerial Image Restoration under Adverse Weather (PLAIR-AW), which aims to recover clean and highquality images from degraded images with bad weather thus improving detection performance for PLAI. In this context, we are the first to release numerous corresponding datasets, namely, HazeCPLID, HazeTTPLA, HazeInsPLAD for power line aerial image dehazing, RainCPLID, RainTTPLA, RainInsPLAD for power line aerial image deraining, SnowCPLID, SnowInsPLAD for power line aerial image desnowing, which are synthesized upon the public power line aerial image datasets of CPLID, TTPLA, InsPLAD following the mathematical models. Meanwhile, we select numerous state-of-the-art methods from image restoration community as the baseline methods for PLAIRAW. At last, we conduct large-scale empirical experiments to evaluate the performance of baseline methods on the proposed datasets. The proposed datasets and trained models are available at https://github.com/ntuhubin/PLAIR-AW