Filters








2,885 Hits in 5.6 sec

An Effective Anti-Aliasing Approach for Residual Networks [article]

Cristina Vasconcelos, Hugo Larochelle, Vincent Dumoulin, Nicolas Le Roux, Ross Goroshin
2020 arXiv   pre-print
Frequency aliasing is a phenomenon that may occur when sub-sampling any signal, such as an image or feature map, causing distortion in the sub-sampled output.  ...  We show that we can mitigate this effect by placing non-trainable blur filters and using smooth activation functions at key locations, particularly where networks lack the capacity to learn them.  ...  Residual Networks.  ... 
arXiv:2011.10675v1 fatcat:ogzwfs75jbdz3objmxjpmw5sdq

GraSens: A Gabor Residual Anti-aliasing Sensing Framework for Action Recognition using WiFi [article]

Yanling Hao, Zhiyuan Shi, Xidong Mu, Yuanwei Liu
2022 arXiv   pre-print
To remedy this issue, we propose an end-to-end Gabor residual anti-aliasing sensing network (GraSens) to directly recognize the actions using the WiFi signals from the wireless devices in diverse scenarios  ...  In each block, the Gabor layer is integrated with the anti-aliasing layer in a residual manner to gain the shift-invariant features.  ...  To remedy the above limitations, in this paper, an end-toend Gabor residual anti-aliasing sensing (GraSens) network is proposed for HAR in varied environments.  ... 
arXiv:2205.11945v1 fatcat:g3437a5fvzctlhggy7nnky46ti

LapEPI-Net: A Laplacian Pyramid EPI structure for Learning-based Dense Light Field Reconstruction [article]

Gaochang Wu and Yebin Liu and Lu Fang and Tianyou Chai
2019 arXiv   pre-print
Accordingly, we design a Laplacian Pyramid EPI (LapEPI) structure that contains both low spatial scale EPI (for aliasing) and high-frequency residuals (for blurring) to solve the trade-off problem.  ...  We then propose a novel network architecture for the LapEPI structure, termed as LapEPI-net.  ...  To tackle the aliasing effect, we adopt pre-filters with different kernel size for the residual pyramid levels.  ... 
arXiv:1902.06221v1 fatcat:55oujzv4cbh2jbaeywu7jfv574

Impact of Aliasing on Generalization in Deep Convolutional Networks [article]

Cristina Vasconcelos, Hugo Larochelle, Vincent Dumoulin, Rob Romijnders, Nicolas Le Roux, Ross Goroshin
2021 arXiv   pre-print
We show how to mitigate aliasing by inserting non-trainable low-pass filters at key locations, particularly where networks lack the capacity to learn them.  ...  We investigate the impact of aliasing on generalization in Deep Convolutional Networks and show that data augmentation schemes alone are unable to prevent it due to structural limitations in widely used  ...  Their anti-aliasing module augments the model with new trainable and non-linear components, increasing the capacity of the network, thus making it unsuitable for isolating the effects of aliasing from  ... 
arXiv:2108.03489v1 fatcat:7h5bld42zzhnzjbtwdmcueidqm

Truly shift-invariant convolutional neural networks [article]

Anadi Chaman
2021 arXiv   pre-print
The existing solutions rely either on data augmentation or on anti-aliasing, both of which have limitations and neither of which enables perfect shift invariance.  ...  Thanks to the use of convolution and pooling layers, convolutional neural networks were for a long time thought to be shift-invariant.  ...  Fig. 5 shows that unlike the baseline and anti-aliasing based approaches, the validation consistency for APS is 100% throughout training.  ... 
arXiv:2011.14214v4 fatcat:h55tijksxza25cnl7zqbnpupkq

How Convolutional Neural Networks Deal with Aliasing [article]

Antônio H. Ribeiro, Thomas B. Schön
2021 arXiv   pre-print
The convolutional neural network (CNN) remains an essential tool in solving computer vision problems.  ...  important role in succeeding at the task; In the second, we show that an image classifier CNN while, in principle, capable of implementing anti-aliasing filters, does not prevent aliasing from taking  ...  For instance, in the context of image classification, Zhang [4] presented a successful implementation of a CNN with anti-aliasing filters, showing that such an approach yields CNNs invariant to shifts  ... 
arXiv:2102.07757v1 fatcat:oaoff4s3tzapjbvsqim7jzv2la

Revisiting Light Field Rendering with Deep Anti-Aliasing Neural Network [article]

Gaochang Wu, Yebin Liu, Lu Fang, Tianyou Chai
2021 arXiv   pre-print
To explore the full potential, we then embed the anti-aliasing framework into a deep neural network through the design of an integrated architecture and trainable parameters.  ...  Instead, we introduce an alternative framework to perform anti-aliasing reconstruction in the image domain and analytically show comparable efficacy on the aliasing issue.  ...  We further introduce an effective framework containing shearing, downscaling and prefiltering operations for anti-aliasing LF reconstruction (Sec. 3.2).  ... 
arXiv:2104.06797v1 fatcat:zvmryixwdfb2jebtphmouhmdgy

Strategies to mitigate aliasing of loading signals while estimating GPS frame parameters

Xavier Collilieux, Tonie van Dam, Jim Ray, David Coulot, Laurent Métivier, Zuheir Altamimi
2011 Journal of Geodesy  
We confirm that the standard approach consisting of estimating the transformation over the whole network is particularly harmful for the loading signals if the network is not well distributed.  ...  We discuss in this paper some procedures that may allow reducing these aliasing effects in the case of the GPS techniques.  ...  The bimodal distribution noticed in the East component can be explained by an aliasing in the X-translation for the out-of-phase term and an aliasing of the X-translation and rotations for the in-phase  ... 
doi:10.1007/s00190-011-0487-6 fatcat:i2smldbxlvflzaog2pecstnxtu

Anti-Aliasing Add-On for Deep Prior Seismic Data Interpolation [article]

Francesco Picetti, Vincenzo Lipari, Paolo Bestagini, Stefano Tubaro
2021 arXiv   pre-print
Among machine learning techniques recently proposed to solve data interpolation as an inverse problem, Deep Prior paradigm aims at employing a convolutional neural network to capture priors on the data  ...  However, this technique lacks of reconstruction precision when interpolating highly decimated data due to the presence of aliasing.  ...  This effect is particularly visible also in the spectra depicted in Figure 2d , where we can notice a residual aliasing pattern; on the other hand, in Figure 2c it has been removed.  ... 
arXiv:2101.11361v1 fatcat:5yaqrwqeenavxgkdh7cq4f2ufu

Exploring Novel Pooling Strategies for Edge Preserved Feature Maps in Convolutional Neural Networks [article]

Adithya Sineesh, Mahesh Raveendranatha Panicker
2021 arXiv   pre-print
With the introduction of anti-aliased convolutional neural networks (CNN), there has been some resurgence in relooking the way pooling is done in CNNs.  ...  In this work, an exhaustive analysis of the edge preserving pooling options for classification, segmentation and autoencoders are presented.  ...  This is an originally submitted version and has not been reviewed by independent peers Jigar Halani, Dr. Manish Modani, Mr. Megh Makwana and Mr.  ... 
arXiv:2110.08842v1 fatcat:vkgs3x5ji5aptfesxcm6dlkbci

Compounding the Performance Improvements of Assembled Techniques in a Convolutional Neural Network [article]

Jungkyu Lee, Taeryun Won, Tae Kwan Lee, Hyemin Lee, Geonmo Gu, Kiho Hong
2020 arXiv   pre-print
Recent studies in image classification have demonstrated a variety of techniques for improving the performance of Convolutional Neural Networks (CNNs).  ...  Our approach achieved 1st place in the iFood Competition Fine-Grained Visual Recognition at CVPR 2019, and the source code and trained models are available at https://github.com/clovaai/assembled-cnn  ...  Down Sampling Block (+Anti Aliasing) Stage 3 Down Sampling Block (+Anti Aliasing) Stage 2 Down Sampling Block Stage 1 Down Sampling Block (+Anti Aliasing) Stage 4 Residual Blocks  ... 
arXiv:2001.06268v2 fatcat:vbv4vakurncjvgvjbd2lbauizm

Learning Spatio-Temporal Downsampling for Effective Video Upscaling [article]

Xiaoyu Xiang, Yapeng Tian, Vijay Rengarajan, Lucas Young, Bo Zhu, Rakesh Ranjan
2022 arXiv   pre-print
Improper spatio-temporal downsampling applied on videos can cause aliasing issues such as moir\'e patterns in space and the wagon-wheel effect in time.  ...  Towards this goal, we propose a neural network framework that jointly learns spatio-temporal downsampling and upsampling.  ...  Acknowledgements We thank Hai Wang for the quantitative results of previous SOTA methods, Salma Abdel Magid for feedback on the manuscript, Chakravarty Reddy Alla Chaitanya and Siddharth Bhargav for discussions  ... 
arXiv:2203.08140v1 fatcat:ffuzgscbu5dzbmgdpl6l7u3jvq

Blending Anti-Aliasing into Vision Transformer [article]

Shengju Qian, Hao Shao, Yi Zhu, Mu Li, Jiaya Jia
2021 arXiv   pre-print
In this work, we analyze the uncharted problem of aliasing in vision transformer and explore to incorporate anti-aliasing properties.  ...  However, the discontinuous patch-wise tokenization process implicitly introduces jagged artifacts into attention maps, arising the traditional problem of aliasing for vision transformers.  ...  We also note that anti-aliasing for modern neural networks still remains an open problem.  ... 
arXiv:2110.15156v1 fatcat:yqnlx7yxcrb2lagqurwf3kictm

WaveCycleGAN2: Time-domain Neural Post-filter for Speech Waveform Generation [article]

Kou Tanaka, Hirokazu Kameoka, Takuhiro Kaneko, Nobukatsu Hojo
2019 arXiv   pre-print
The results show that the proposed method 1) alleviates the aliasing well, 2) is useful for both speech waveforms generated by analysis-and-synthesis and statistical parametric speech synthesis, and 3)  ...  One possible cause of this distinguishability is the aliasing observed in the processed speech waveform via down/up-sampling modules.  ...  This is reasonable because the classical convolution with strides is not guaranteed to have an antialiasing mechanism while we never perform down-sampling without anti-aliasing processing in the pure signal  ... 
arXiv:1904.02892v2 fatcat:l3riailo7bhabg4jlwimrf5dvy

Anti-aliasing Semantic Reconstruction for Few-Shot Semantic Segmentation [article]

Binghao Liu and Yao Ding and Jianbin Jiao and Xiangyang Ji and Qixiang Ye
2021 arXiv   pre-print
Our proposed approach, referred to as anti-aliasing semantic reconstruction (ASR), provides a systematic yet interpretable solution for few-shot learning problems.  ...  By introducing contrastive loss, we maximize the orthogonality of basis vectors while minimizing semantic aliasing between classes.  ...  In this paper, we reformulate the few-shot segmentation task as a semantic reconstruction problem and propose an anti-aliasing semantic reconstruction (ASR) approach.  ... 
arXiv:2106.00184v1 fatcat:2dhpjq3h3nbenaill6xjiwwecm
« Previous Showing results 1 — 15 out of 2,885 results