Filters








22,979 Hits in 5.1 sec

Cross-dimensional Weighting for Aggregated Deep Convolutional Features [article]

Yannis Kalantidis, Clayton Mellina, Simon Osindero
2016 arXiv   pre-print
We propose a simple and straightforward way of creating powerful image representations via cross-dimensional weighting and aggregation of deep convolutional neural network layer outputs.  ...  We first present a generalized framework that encompasses a broad family of approaches and includes cross-dimensional pooling and weighting steps.  ...  We base our cross-dimensional weighted features on a generic deep convolutional neural network.  ... 
arXiv:1512.04065v2 fatcat:cklunmdnz5d2hb6cz4pvv2pqjq

Unsupervised Semantic-based Aggregation of Deep Convolutional Features

Jian Xu, Chunheng Wang, Chengzuo Qi, Cunzhao Shi, Baihua Xiao
2018 IEEE Transactions on Image Processing  
In this paper, we propose a simple but effective semantic-based aggregation (SBA) method. The proposed SBA utilizes the discriminative filters of deep convolutional layers as semantic detectors.  ...  The final global SBA representation could then be acquired by aggregating the regional representations weighted by the selected "probabilistic proposals" corresponding to various semantic content.  ...  [13] extend the work of [11] by allowing cross-dimensional weighting.  ... 
doi:10.1109/tip.2018.2867104 pmid:30281422 fatcat:yr224g546fdq5fj2mqxhmlxicq

Adaptive Co-weighting Deep Convolutional Features For Object Retrieval [article]

Jiaxing Wang, Jihua Zhu, Shanmin Pang, Zhongyu Li, Yaochen Li, Xueming Qian
2018 arXiv   pre-print
Aggregating deep convolutional features into a global image vector has attracted sustained attention in image retrieval.  ...  In this paper, we propose an efficient unsupervised aggregation method that uses an adaptive Gaussian filter and an elementvalue sensitive vector to co-weight deep features.  ...  The major difference between them is CroW employs a more effective cross-dimensional weighting strategy to weight deep features.  ... 
arXiv:1803.07360v1 fatcat:txw73mu3abch3aykkcv4mtlgem

ECA-Net: Efficient Channel Attention for Deep Convolutional Neural Networks [article]

Qilong Wang, Banggu Wu, Pengfei Zhu, Peihua Li, Wangmeng Zuo, Qinghua Hu
2020 arXiv   pre-print
Therefore, we propose a local cross-channel interaction strategy without dimensionality reduction, which can be efficiently implemented via 1D convolution.  ...  By dissecting the channel attention module in SENet, we empirically show avoiding dimensionality reduction is important for learning channel attention, and appropriate cross-channel interaction can preserve  ...  GE [13] explores spatial extension using a depth-wise convolution [5] to aggregate features.  ... 
arXiv:1910.03151v4 fatcat:pwclleuo6ff6xpfnn3tcq34qri

Toward Improving Image Retrieval via Global Saliency Weighted Feature

Hongwei Zhao, Jiaxin Wu, Danyang Zhang, Pingping Liu
2021 ISPRS International Journal of Geo-Information  
For full description of images' semantic information, image retrieval tasks are increasingly using deep convolution features trained by neural networks.  ...  However, to form a compact feature representation, the obtained convolutional features must be further aggregated in image retrieval. The quality of aggregation affects retrieval performance.  ...  Data Availability Statement: Data and the optimal network mode for this research work are stored in Github: https://github.com/wujx990/feizai_upload, accessed on 6 April 2021.  ... 
doi:10.3390/ijgi10040249 fatcat:he2q44jfajgpdcff43js4up6va

Deep Aggregation of Regional Convolutional Activations for Content Based Image Retrieval [article]

Konstantin Schall, Kai Uwe Barthel, Nico Hezel, Klaus Jung
2019 arXiv   pre-print
Subsequently, weights for each of the pooled feature vectors are learned to perform a weighted aggregation to a single feature vector.  ...  One of the key challenges of deep learning based image retrieval remains in aggregating convolutional activations into one highly representative feature vector.  ...  For training, the authors further sum-aggregate the weighted combination of the 49 embeddings to form a single 2048-dimensional feature vector. Training is done in two steps.  ... 
arXiv:1909.09420v2 fatcat:v27lbs7wmbbefhjo76k7ftai5i

ECA-Net: Efficient Channel Attention for Deep Convolutional Neural Networks

Qilong Wang, Banggu Wu, Pengfei Zhu, Peihua Li, Wangmeng Zuo, Qinghua Hu
2020 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
By dissecting the channel attention module in SENet, we empirically show avoiding dimensionality reduction is important for learning channel attention, and appropriate cross-channel interaction can preserve  ...  Recently, channel attention mechanism has demonstrated to offer great potential in improving the performance of deep convolutional neural networks (CNNs).  ...  GE [13] explores spatial extension using a depth-wise convolution [5] to aggregate features.  ... 
doi:10.1109/cvpr42600.2020.01155 dblp:conf/cvpr/WangWZLZH20 fatcat:ofokz2igkzfrzgvrao4sbx3ws4

SplineCNN: Fast Geometric Deep Learning with Continuous B-Spline Kernels [article]

Matthias Fey, Jan Eric Lenssen, Frank Weichert, Heinrich Müller
2018 arXiv   pre-print
We present Spline-based Convolutional Neural Networks (SplineCNNs), a variant of deep neural networks for irregular structured and geometric input, e.g., graphs or meshes.  ...  In contrast to related approaches that filter in the spectral domain, the proposed method aggregates features purely in the spatial domain.  ...  We also thank Pascal Libuschewski for proofreading and helpful advice.  ... 
arXiv:1711.08920v2 fatcat:mywhfxmpnfhnfkno2jjsifwv6e

CA-XTree: Age Estimation of Grouped Gradient Regression Tree with Local Channel Attention

Xiaoding Lu, Zhengyou Wang, Yanhui Xia, Shanna Zhuang, Dalin Zhang
2022 Computational Intelligence and Neuroscience  
Firstly, features are extracted through the convolution layer and then combined with the local channel attention module to strengthen the ability of age feature information interaction between different  ...  This paper improves state of the art for image classification on MORPH and CACD datasets. The advantage of our model is that it is easy to implement and has no excess memory overhead.  ...  In this paper, the size k of convolution kernel is adaptively determined by using the global average pooling aggregation of convolution features without dimensionality reduction, and then convolution operation  ... 
doi:10.1155/2022/4155461 pmid:35669653 pmcid:PMC9167079 fatcat:fqksgzcowjcxpameibv6wvvpjq

Inverted Residual Siamese Visual Tracking with Feature Crossing Network

Feng Zhang, Xiaoyan Qian, Lei Han, Yi Shen
2021 IEEE Access  
Feature-crossing network is to perform feature-level aggregations, which makes deep and shallow layers complement each other more closely and further improves tracking accuracy.  ...  Specifically, the Siamese backbone networks for feature extraction consist of an inverted residual network and a feature-crossing network (FCN).  ...  As shown in Fig.1 , this proposed framework consists of an inverted residual network for feature extraction, a cross fusion network for feature-level aggregation and three region proposed subnetworks  ... 
doi:10.1109/access.2021.3056194 fatcat:gkne5xwwpjglvas3r3g5i3hqqe

Wireless Network Intrusion Detection Based on Improved Convolutional Neural Network

Hongyu Yang, Fengyan Wang
2019 IEEE Access  
The low-level intrusion traffic data is abstractly represented as advanced features by CNN, which extracted autonomously the sample features, and optimizing network parameters by stochastic gradient descent  ...  INDEX TERMS Wireless network intrusion detection, security, convolutional neural network. 64366 2169-3536  ...  Each convolution kernel corresponds to a specific feature map for network weight learning.  ... 
doi:10.1109/access.2019.2917299 fatcat:6zr4af2xlncozg4rol256v37ba

Multilayer Convolutional Feature Aggregation Algorithm for Image Retrieval

Rongsheng Dong, Ming Liu, Fengying Li
2019 Mathematical Problems in Engineering  
A new image description algorithm ML-RCroW based on multilayer multiregion cross-weighted aggregational deep convolutional features is proposed.  ...  Finally, features of each layer of the VGG16 network model are extracted and then aggregated and dimensionally reduced to obtain the final feature vector of the image.  ...  In this paper, a multilayer multiregion cross-weighted matrix aggregation deep convolutional feature (ML-RCroW) algorithm is proposed for image retrieval based on the literature [16] .  ... 
doi:10.1155/2019/9794202 fatcat:fi746ksiuzhlpjefeufybymhdu

SplineCNN: Fast Geometric Deep Learning with Continuous B-Spline Kernels

Matthias Fey, Jan Eric Lenssen, Frank Weichert, Heinrich Muller
2018 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition  
In contrast to related approaches that filter in the spectral domain, the proposed method aggregates features purely in the spatial domain.  ...  We present Spline-based Convolutional Neural Networks (SplineCNNs), a variant of deep neural networks for irregular structured data.  ...  We also thank Pascal Libuschewski for proofreading and helpful advice.  ... 
doi:10.1109/cvpr.2018.00097 dblp:conf/cvpr/FeyLWM18 fatcat:w25ozz3fqvg3lk6jygrzlgaplm

Integrating Normal Vector Features into an Atrous Convolution Residual Network for LiDAR Point Cloud Classification

Chunjiao Zhang, Shenghua Xu, Tao Jiang, Jiping Liu, Zhengjun Liu, An Luo, Yu Ma
2021 Remote Sensing  
The improved local feature aggregation module can merge the deep features of the point cloud and mine the fine-grained information of the point cloud to improve the model's segmentation ability in complex  ...  Then, the point cloud normal vector is embedded in the local feature aggregation module of the RandLA-Net network to extract local semantic aggregation features.  ...  Through the attention score, the essential features with high scores are selected. Finally, these features are aggregated after weighted summation.  ... 
doi:10.3390/rs13173427 fatcat:6w6a6l3zlzex3bjyoflqks2njq

An Information-rich Sampling Technique over Spatio-Temporal CNN for Classification of Human Actions in Videos [article]

S.H. Shabbeer Basha, Viswanath Pulabaigari, Snehasis Mukherjee
2020 arXiv   pre-print
We propose a novel scheme for human action recognition in videos, using a 3-dimensional Convolutional Neural Network (3D CNN) based classifier.  ...  In the proposed video sampling technique, consecutive k frames of a video are aggregated into a single frame by computing a Gaussian-weighted summation of the k frames.  ...  ACKNOWLEDGMENTS We acknowledge the support of NVIDIA with the donation of the GeForce Titan XP GPU used for this research.  ... 
arXiv:2002.02100v2 fatcat:jqk7s65wazcr3ocwfy7qe3agyq
« Previous Showing results 1 — 15 out of 22,979 results