检索结果-内蒙古大学图书馆

Sasl: Saliency-adaptive sparsity learning for neural network acceleration

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Shi, Jun Xu, Jianfeng Tasaka, Kazuyuki Chen, Zhibo CAS Key Laboratory of Technology in Geo-spatial Information Processing and Application System University of Science and Technology of China Hefei230027 China KDDI Research Japan

Accelerating the inference speed of CNNs is critical to their deployment in real-world applications. Among all the pruning approaches, those implementing a sparsity learning framework have shown to be effective as they learn and prune the models in an end-to-end data-driven manner. However, these works impose the same sparsity regularization on all filters indiscriminately, which can hardly result in an optimal structuresparse network. In this paper, we propose a Saliency-Adaptive Sparsity Learning (SASL) approach for further optimization. A novel and effective estimation of each filter, i.e., saliency, is designed, which is measured from two aspects: The importance for the prediction performance and the consumed computational resources. During sparsity learning, the regularization strength is adjusted according to the saliency, so our optimized format can better preserve the prediction performance while zeroing out more computation-heavy filters. The calculation for saliency introduces minimum overhead to the training process, which means our SASL is very efficient. During the pruning phase, in order to optimize the proposed data-dependent criterion, a hard sample mining strategy is utilized, which shows higher effectiveness and efficiency. Extensive experiments demonstrate the superior performance of our method. Notably, on ILSVRC-2012 dataset, our approach can reduce 49.7% FLOPs of ResNet-50 with very negligible 0.39% top-1 and 0.05% top-5 accuracy degradation. Copyright © 2020, The Authors. All rights reserved.

关键词： Convolutional neural networks

DEEP UNFOLDING NETWORK FOR SPARSE SAR IMAGING BASED ON COMPOUND REGULARIZATION

学校读者我要写书评

暂无评论

DEEP UNFOLDING NETWORK FOR SPARSE SAR IMAGING BASED ON COMPO...

IET International Radar Conference 2023, IRC 2023

作者： Zhou, Guoru Xu, Zhongqiu Fan, Yizhe Zhang, Zhe Zhang, Bingchen Wu, Yirong Aerospace Information Research Institute Chinese Academy of Sciences Beijing China Key Laboratory of Technology in Geo-spatial Information Processing and Application System Chinese Academy of Sciences Beijing China School of Electronic Electrical and Communication Engineering University of Chinese Academy of Sciences Beijing China

ISBN: (纸本)9781839539954

How to get an accurate image of multi-class targets scene is the key to synthetic aperture radar (SAR) imaging research. In sparse SAR imaging, compound regularization constrains multi-class target features by setting multiple regularizations prior to the model. However, the solution method based on a traditional iterative algorithm needs to set parameters manually and has a large number of iterations. In this paper, a novel SAR imaging method for the l1 norm and the total variation (TV) compound regularization based on a deep unfolding network (DUN) is proposed. The l1 norm and TV compound regularization can promote the sparsity and despeckle at the same time. Besides, the deep unfolded network can learn iteration parameters in a data-driven manner to speed up convergence. Finally, real SAR data experiments can verify the effectiveness of the proposed method. © The Institution of Engineering & technology 2023.

关键词： Radar imaging

Offline and Online Optical Flow Enhancement for Deep Video Compression

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Tang, Chuanbo Sheng, Xihua Li, Zhuoyuan Zhang, Haotian Li, Li Liu, Dong CAS Key Laboratory of Technology in Geo-spatial Information Processing and Application System Department of Electronic Engineering and Information Science University of Science and Technology of China Hefei230027 China

Video compression relies heavily on exploiting the temporal redundancy between video frames, which is usually achieved by estimating and using the motion information. The motion information is represented as optical flows in most of the existing deep video compression networks. Indeed, these networks often adopt pre-trained optical flow estimation networks for motion estimation. The optical flows, however, may be less suitable for video compression due to the following two factors. First, the optical flow estimation networks were trained to perform inter-frame prediction as accurately as possible, but the optical flows themselves may cost too many bits to encode. Second, the optical flow estimation networks were trained on synthetic data, and may not generalize well enough to real-world videos. We address the twofold limitations by enhancing the optical flows in two stages: offline and online. In the offline stage, we fine-tune a trained optical flow estimation network with the motion information provided by a traditional (non-deep) video compression scheme, e.g. H.266/VVC, as we believe the motion information of H.266/VVC achieves a better rate-distortion tradeoff. In the online stage, we further optimize the latent features of the optical flows with a gradient descent-based algorithm for the video to be compressed, so as to enhance the adaptivity of the optical flows. We conduct experiments on a state-of-the-art deep video compression scheme, DCVC. Experimental results demonstrate that the proposed offline and online enhancement together achieves on average 12.8% bitrate saving on the tested videos, without increasing the model or computational complexity of the decoder side. Copyright © 2023, The Authors. All rights reserved.

关键词： Image coding

Image-to-image translation with multi-path consistency regularization

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Lin, Jianxin Xia, Yingce Wang, Yijun Qin, Tao Chen, Zhibo CAS Key Laboratory of Technology in Geo-spatial Information Processing and Application System University of Science and Technology of China University of Science and Technology of China Microsoft Research Asia

Image translation across different domains has attracted much attention in both machine learning and computer vision communities. Taking the translation from source domain Ds to target domain Dt as an example, existing algorithms mainly rely on two kinds of loss for training: One is the discrimination loss, which is used to differentiate images generated by the models and natural images;the other is the reconstruction loss, which measures the difference between an original image and the reconstructed version through Ds → Dt → Ds translation. In this work, we introduce a new kind of loss, multi-path consistency loss, which evaluates the differences between direct translation Ds → Dt and indirect translation Ds → Da → Dt with Da as an auxiliary domain, to regularize training. For multi-domain translation (at least, three) which focuses on building translation models between any two domains, at each training iteration, we randomly select three domains, set them respectively as the source, auxiliary and target domains, build the multi-path consistency loss and optimize the network. For two-domain translation, we need to introduce an additional auxiliary domain and construct the multi-path consistency loss. We conduct various experiments to demonstrate the effectiveness of our proposed methods, including face-toface translation, paint-to-photo translation, and deraining/de-noising translation. Copyright © 2019, The Authors. All rights reserved.

关键词： Iterative methods

ISAR imaging based on antenna array and compressed sensing

学校读者我要写书评

暂无评论

ISAR imaging based on antenna array and compressed sensing

6th International Conference on Radar, RADAR 2011

作者： Zhong, Li-Hua Hu, Dong-Hui Ding, Chi-Biao Institute of Electronics Chinese Academy of Sciences Beijing 100190 China Graduate University Chinese Academy of Sciences Beijing 100039 China Key Laboratory of Technology in Geo-spatial Information Processing and Application System Institute of Electronics Chinese Academy of Sciences 100190 China

ISBN: (纸本)9781424484416

Compressed sensing (CS) has recently attracted much interest for its ability to recovery a sparse signal with very limited number of samples. In this paper, we adapt this idea and present a framework of high-resolution ISAR imaging based on antenna array. During the framework, linear antenna array is analyzed and compared with different conventional algorithms. Numerical simulation results confirmed the validity of the proposed algorithm. © 2011 IEEE.

关键词： Radar imaging

Image Deblocking via Group Sparsity Optimization

学校读者我要写书评

暂无评论

Image Deblocking via Group Sparsity Optimization

IEEE International Symposium on Circuits and systems

作者： Zhenbo Lu Houqiang Li Weiping Li CAS Key Laboratory of Technology in Geo-spatial Information Processing and Application System Department of Electronic Engineering and Information Science University of Science and Technology of China Heifei Anhui China

ISBN: (纸本)9781479983926

Block-wise compressed image often suffers from the blocking artifacts. In this paper, we propose a novel deblocking scheme for compressed image, by combining image's sparse property and its self-similarity together, called group sparsity optimization. Instead of processing each image patch individually, in the proposed scheme, similar patches in one group are required to be well-represented on learned dictionary collaboratively, using group sparsity regularization. The group sparsity not only imposes every patch's representation to be sparse, bus also requires patches' coefficients in the group share the similar pattern. The experiment results on standard test images demonstrate that our scheme can improve the PSNR of the compressed images by an average of 1.25 dB, and outperform state of the art deblocking approaches.

关键词： Image Deblocking Image Restoration JPEG compression

Improve spaceborne SAR geometric accuracy using DORIS

学校读者我要写书评

暂无评论

Improve spaceborne SAR geometric accuracy using DORIS

IET International Radar Conference 2009

作者： Jiayin, Liu Wen, Hong Yueguan, Lin Key Laboratory of Technology in Geo-spatial Information Processing and Application System Institute of Electronics Chinese Academy of Sciences Beijing 100190 China National Key Laboratory of Microwave Imaging Technology Institute of Electronics Chinese Academy of Sciences Beijing 100190 China Graduate University Chinese Academy of Sciences Beijing 100190 China

ISBN: (纸本)9781849190107

The paper analyzes the effects of the orbit position errors in terms of the geometric accuracy of the space-borne SAR based on the range-Doppler method. The paper gives the simulation experiment results of the space-borne SAR geometric accuracy using DORIS, GPS and MADRAS. The results show the space-borne SAR geometric accuracy is 1 meter using DORIS, 80 meter higher than using MADRAS, and 50 meter higher than using GPS.

关键词： geometry

Electromagnetic Simulation Aided SAR Target Classification Via Deep Domain Adaptation

学校读者我要写书评

暂无评论

Electromagnetic Simulation Aided SAR Target Classification V...

2021 CIE International Conference on Radar, Radar 2021

作者： Lv, Xiaoling Qiu, Xiaolan Yu, Wenming Aerospace Information Research Institute Chinese Academy of Sciences School of Electronic Electrical and Communication Engineering University of Chinese Academy of Sciences Beijing China Suzhou Aerospace Information Research Institute Key Laboratory of Technology in Geo-Spatial Information Processing and Application System Suzhou China Southeast University State Key Laboratory of Millimeter Waves Nanjing China

ISBN: (纸本)9781665498142

Convolutional neural networks (CNNs) have made tremendous success in optical images classification recently. However, in synthetic aperture radar (SAR) target classification, it is difficult to annotate a large amount of real SAR images to train CNNs. Sufficient annotated images can be easily obtained through simulation, but the disparity between the simulated images and the real images makes them difficult to directly apply to the real images classification. In this paper, we propose a model that integrates multi-kernel maximum mean discrepancy (MK-MMD) and domain-adversarial training to alleviate this problem. Simulated SAR images with annotation and unlabeled real SAR images are used to train our model. First, we use domain-adversarial training to prompt the model to extract domain-invariant features. Then, the MK-MMD between the hidden representations of simulated images and real images is reduced to narrow domain discrepancy. Experimental results on the real SAR dataset demonstrate that our method effectively solves the domain shift problem and improves the classification accuracy. © 2021 IEEE.

关键词： convolutional neural network (CNN) Synthetic aperture radar (SAR) target classification unsupervised domain adaptation (UDA)

Towards a better match in siamese network based visual object tracker

学校读者我要写书评

暂无评论

arXiv 2018年

作者： He, Anfeng Luo, Chong Tian, Xinmei Zeng, Wenjun CAS Key Laboratory of Technology in Geo-Spatial Information Processing and Application System University of Science and Technology of China Hefei Anhui China Microsoft Research Beijing China

Recently, Siamese network based trackers have received tremendous interest for their fast tracking speed and high performance. Despite the great success, this tracking framework still suffers from several limitations. First, it cannot properly handle large object rotation. Second, tracking gets easily distracted when the background contains salient objects. In this paper, we propose two simple yet effective mechanisms, namely angle estimation and spatial masking, to address these issues. The objective is to extract more representative features so that a better match can be obtained between the same object from different frames. The resulting tracker, named Siam-BM, not only significantly improves the tracking performance, but more importantly maintains the realtime capability. Evaluations on the VOT2017 dataset show that Siam-BM achieves an EAO of 0.335, which makes it the best-performing realtime tracker to date. Copyright © 2018, The Authors. All rights reserved.

关键词： Deep neural networks