Yuzhen Shi, Huanghai Liu, Yiran HU, Song Gaojie, Xu Xinran, Yubo Ma, Tianyi Tang, Li Zhang, Qingjing Chen, Feng Di, Wenbo Lv, Weiheng Wu, Kexin Yang, Sen Yang, Wei Wang, Rongyao Shi, Qiu Yuanyang, Yuemeng Qi, Zhang Jingwen, Sui Xiaoyu, Yifan Chen, Zhang Yi, An Yang, Bowen Yu, Dayiheng Liu, Junyang Lin, Weixing Shen, Bing Zhao, Charles L. A. Clarke, HU Wei. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Atrous convolutions are employed as a method to increase the receptive field in semantic segmentation tasks. However, in previous works of semantic segmentation, it was rarely employed in the shallow layers of the model. We revisit the design of atrous convolutions in modern convolutional neural networks (CNNs), and demonstrate that the concept of using large kernels to apply atrous convolutions could be a more powerful paradigm. We propose three guidelines to apply atrous convolutions more efficiently: Do not only use atrous convolutions, Avoiding the “Atrous Disasters”, Appropriate fusion mechanisms make it perfect. Following these guidelines, we propose DSNet, a Dual-Branch CNN architecture, which incorporates atrous convolutions in the shallow layers of the model architecture, as well as pretraining the nearly entire encoder on ImageNet to achieve better performance. To demonstrate the effectiveness of our approach, our models achieve a new state-of-the-art trade-off between accuracy and speed on ADE20K, Cityscapes and BDD datasets. Specifically, DSNet achieves 40.0% mIOU with inference speed of 179.2 FPS on ADE20K, and 80.4% mIOU with speed of 81.9 FPS on Cityscapes. Additionally, we propose a novel multi-scale attention fusion module, MSAF. It demonstrates outstanding performance in classification as well as downstream tasks such as segmentation. Source code and models are available at Github:https://github.com/takaniwa/DSNet.

Coarse-to-fine schemes are widely used in traditional single-image motion deblur; however, in the context of deep learning, existing multi-scale algorithms not only require the use of complex modules for feature fusion of low-scale RGB images and deep semantics, but also manually generate low-resolution pairs of images that do not have sufficient confidence. In this work, we propose a multi-scale network based on single-input and multiple-outputs(SIMO) for motion deblurring. This simplifies the complexity of algorithms based on a coarse-to-fine scheme. To alleviate restoration defects impacting detail information brought about by using a multi-scale architecture, we combine the characteristics of real-world blurring trajectories with a learnable wavelet transform module to focus on the directional continuity and frequency features of the step-by-step transitions between blurred images to sharp images. In conclusion, we propose a multi-scale network with a learnable discrete wavelet transform (MLWNet), which exhibits state-of-the-art performance on multiple real-world deblurred datasets, in terms of both subjective and objective quality as well as computational efficiency. Our code is available on https://github.com/thqiu0419/MLWNet.

SimRank is an important measure of vertex-pair similarity according to the structure of graphs. Although progress has been achieved, existing methods still face challenges to handle large graphs. Besides huge index construction and maintenance cost, existing methods may require considerable search space and time overheads in the online SimRank query. In this paper, we design a Monte Carlo based method, UniWalk, to enable the fast top-k SimRank computation over large undirected graphs. UniWalk directly locates the top-k similar vertices for any single source vertex u via R sampling paths originating from u, which avoids selecting candidate vertex set C and the following O(1C1R) bidirectional sampling paths. We also devise a path enumeration strategy to improve the SimRank precision by using path probabilities instead of path frequencies when sampling, a space-efficient method to reduce intermediate results, and a path-sharing strategy to lower the redundant path sampling cost for multiple source vertices. Furthermore, we extend UniWalk to existing distributed graph processing frameworks to improve its scalability. We conduct extensive experiments to illustrate that UniWalk has high scalability, and outperforms the state-of-the-art methods by orders of magnitude.

In this paper, we study the One-Sender-Multiple-Receiver (OSMR) transmission technique, which allows one sender to send to multiple receivers simultaneously by utilizing multiple antennas at the sender. To study the physical-layer characteristics of OSMR, we implement a prototype OSMR transmitter/receiver with GNU software defined radio and conduct experiments in a university building. Our results are positive and show that wireless channels allow OSMR for a significant percentage of the time. Motivated by our physical-layer study, we propose extensions to the 802.11 MAC protocol to support OSMR transmission, which is backward-compatible with existing 802.11 devices. We also note that the access point (AP) needs a packet scheduling algorithm to efficiently exploit OSMR. We show that the scheduling problem without considering the packet transmission overhead can be formalized as a linear programming problem, but the scheduling problem considering the overhead is NP-hard. We then propose a practical scheduler based on a two-phase algorithm that can also handle channel fluctuations. We test the proposed protocol and algorithm with simulations driven by traffic traces collected from wireless LANs and channel-state traces collected from our experiments, and the results show that OSMR significantly improves the downlink performance.

The paper presents the architecture of an unmanned vehicle called `Intelligent Pioneer' developed by the authors. The unmanned vehicle has the ability of navigating in urban environments autonomously without GPS or any other satellite navigation system. The vehicle is able to select its own routes by perceive the traffic signs and interact with other vehicles or pedestrians under the traffic rules. Various urban driving skills including lane keeping, U-turns, parking, obstacle avoidance, and merging into moving traffic have been developed for the vehicle. The vehicle attended the Future Challenge of Intelligent Vehicles in China organized by National Nature Science Foundation of China in 2010 and 2011. `Intelligent Pioneer' finished all of the competition programs and won the first position in 2010 and the third position in 2011.

In this paper, we study the One-Sender-Multiple-Receiver (OSMR) transmission technique, which allows one sender to send to multiple receivers simultaneously by utilizing multiple antennas at the sender. We implemented a prototype OSMR transmitter/receiver with GNU Software Defined Radio, and conducted experiments in a university building to study the physical layer characteristics of OSMR. Our results are positive and show that wireless channels allow OSMR for a significant percentage of the time. Motivated by our physical layer study, we propose extensions to the 802.11 MAC protocol to support OSMR transmission, which is backward compatible with existing 802.11 devices. We also note that the AP needs a packet scheduling algorithm to efficiently exploit OSMR. We show that the scheduling problem without considering the packet transmission overhead can be formalized as a Linear Programming problem, but the scheduling problem considering the overhead is NP-hard. We then propose a greedy algorithm to schedule OSMR transmissions. We tested the proposed protocol and algorithm with simulations driven by traffic traces collected from wireless LANs and channel state traces collected from our experiments, and the results show that OSMR significantly improves the downlink performance.

In H.264, Intra prediction is employed to reduce energy of prediction error to be transmitted. Because the order of Intra frame encoding is from left to right horizontally and from top to bottom vertically, intra prediction reference pels can only appear on the left side or upside of the block to be predicted, so predicted results are not ideal sometimes. In this paper, a new intra 4 times 4 prediction algorithm based on flexible order of sub blocks encoding is introduced, which improved the modes of intra prediction and the set of reference pels for prediction. Experimental results show that compared to the standard intra prediction mode of H.264, this proposed method can gain 0.4 dBs of PSNR averagely.