Convolutional Neural Networks Without Any Checkerboard Artifacts. Sugawara, Y., Shiota, S., & Kiya, H. In 2018 26th European Signal Processing Conference (EUSIPCO), pages 1317-1321, Sep., 2018.
![Convolutional Neural Networks Without Any Checkerboard Artifacts [pdf]](https://bibbase.org/img/filetypes/pdf.svg "pdf")
Paper doi abstract bibtex
Paper doi abstract bibtex It is well-known that a number of convolutional neural networks (CNNs) generate checkerboard artifacts in both of two processes: forward-propagation of upsampling layers and backpropagation of convolutional layers. A condition to avoid the checkerboard artifacts is proposed in this paper. So far, checkerboard artifacts have been mainly studied for linear multirate systems, but the condition to avoid checkerboard artifacts can not be applied to CNNs due to the non-linearity of CNNs. We extend the avoiding condition for CNNs, and apply the proposed structure to some typical CNNs to confirm the effectiveness of the new scheme. Experiment results demonstrate that the proposed structure can perfectly avoid to generate checkerboard artifacts, while keeping excellent properties that the CNNs have.
@InProceedings{8553099,
author = {Y. Sugawara and S. Shiota and H. Kiya},
booktitle = {2018 26th European Signal Processing Conference (EUSIPCO)},
title = {Convolutional Neural Networks Without Any Checkerboard Artifacts},
year = {2018},
pages = {1317-1321},
abstract = {It is well-known that a number of convolutional neural networks (CNNs) generate checkerboard artifacts in both of two processes: forward-propagation of upsampling layers and backpropagation of convolutional layers. A condition to avoid the checkerboard artifacts is proposed in this paper. So far, checkerboard artifacts have been mainly studied for linear multirate systems, but the condition to avoid checkerboard artifacts can not be applied to CNNs due to the non-linearity of CNNs. We extend the avoiding condition for CNNs, and apply the proposed structure to some typical CNNs to confirm the effectiveness of the new scheme. Experiment results demonstrate that the proposed structure can perfectly avoid to generate checkerboard artifacts, while keeping excellent properties that the CNNs have.},
keywords = {backpropagation;feedforward neural nets;convolutional neural networks;checkerboard artifacts;linear multirate systems;upsampling layer forward-propagation;CNN;Convolution;Deconvolution;Image resolution;Linear systems;Signal resolution;Steady-state;Europe;Convolutional Neural Networks;Checkerboard Artifacts},
doi = {10.23919/EUSIPCO.2018.8553099},
issn = {2076-1465},
month = {Sep.},
url = {https://www.eurasip.org/proceedings/eusipco/eusipco2018/papers/1570436576.pdf},
}Downloads: 0
{"_id":"GSNwRbdkn5n2dopP7","bibbaseid":"sugawara-shiota-kiya-convolutionalneuralnetworkswithoutanycheckerboardartifacts-2018","authorIDs":[],"author_short":["Sugawara, Y.","Shiota, S.","Kiya, H."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","author":[{"firstnames":["Y."],"propositions":[],"lastnames":["Sugawara"],"suffixes":[]},{"firstnames":["S."],"propositions":[],"lastnames":["Shiota"],"suffixes":[]},{"firstnames":["H."],"propositions":[],"lastnames":["Kiya"],"suffixes":[]}],"booktitle":"2018 26th European Signal Processing Conference (EUSIPCO)","title":"Convolutional Neural Networks Without Any Checkerboard Artifacts","year":"2018","pages":"1317-1321","abstract":"It is well-known that a number of convolutional neural networks (CNNs) generate checkerboard artifacts in both of two processes: forward-propagation of upsampling layers and backpropagation of convolutional layers. A condition to avoid the checkerboard artifacts is proposed in this paper. So far, checkerboard artifacts have been mainly studied for linear multirate systems, but the condition to avoid checkerboard artifacts can not be applied to CNNs due to the non-linearity of CNNs. We extend the avoiding condition for CNNs, and apply the proposed structure to some typical CNNs to confirm the effectiveness of the new scheme. Experiment results demonstrate that the proposed structure can perfectly avoid to generate checkerboard artifacts, while keeping excellent properties that the CNNs have.","keywords":"backpropagation;feedforward neural nets;convolutional neural networks;checkerboard artifacts;linear multirate systems;upsampling layer forward-propagation;CNN;Convolution;Deconvolution;Image resolution;Linear systems;Signal resolution;Steady-state;Europe;Convolutional Neural Networks;Checkerboard Artifacts","doi":"10.23919/EUSIPCO.2018.8553099","issn":"2076-1465","month":"Sep.","url":"https://www.eurasip.org/proceedings/eusipco/eusipco2018/papers/1570436576.pdf","bibtex":"@InProceedings{8553099,\n author = {Y. Sugawara and S. Shiota and H. Kiya},\n booktitle = {2018 26th European Signal Processing Conference (EUSIPCO)},\n title = {Convolutional Neural Networks Without Any Checkerboard Artifacts},\n year = {2018},\n pages = {1317-1321},\n abstract = {It is well-known that a number of convolutional neural networks (CNNs) generate checkerboard artifacts in both of two processes: forward-propagation of upsampling layers and backpropagation of convolutional layers. A condition to avoid the checkerboard artifacts is proposed in this paper. So far, checkerboard artifacts have been mainly studied for linear multirate systems, but the condition to avoid checkerboard artifacts can not be applied to CNNs due to the non-linearity of CNNs. We extend the avoiding condition for CNNs, and apply the proposed structure to some typical CNNs to confirm the effectiveness of the new scheme. Experiment results demonstrate that the proposed structure can perfectly avoid to generate checkerboard artifacts, while keeping excellent properties that the CNNs have.},\n keywords = {backpropagation;feedforward neural nets;convolutional neural networks;checkerboard artifacts;linear multirate systems;upsampling layer forward-propagation;CNN;Convolution;Deconvolution;Image resolution;Linear systems;Signal resolution;Steady-state;Europe;Convolutional Neural Networks;Checkerboard Artifacts},\n doi = {10.23919/EUSIPCO.2018.8553099},\n issn = {2076-1465},\n month = {Sep.},\n url = {https://www.eurasip.org/proceedings/eusipco/eusipco2018/papers/1570436576.pdf},\n}\n\n","author_short":["Sugawara, Y.","Shiota, S.","Kiya, H."],"key":"8553099","id":"8553099","bibbaseid":"sugawara-shiota-kiya-convolutionalneuralnetworkswithoutanycheckerboardartifacts-2018","role":"author","urls":{"Paper":"https://www.eurasip.org/proceedings/eusipco/eusipco2018/papers/1570436576.pdf"},"keyword":["backpropagation;feedforward neural nets;convolutional neural networks;checkerboard artifacts;linear multirate systems;upsampling layer forward-propagation;CNN;Convolution;Deconvolution;Image resolution;Linear systems;Signal resolution;Steady-state;Europe;Convolutional Neural Networks;Checkerboard Artifacts"],"metadata":{"authorlinks":{}},"downloads":0},"bibtype":"inproceedings","biburl":"https://raw.githubusercontent.com/Roznn/EUSIPCO/main/eusipco2018url.bib","creationDate":"2021-02-13T15:38:40.121Z","downloads":0,"keywords":["backpropagation;feedforward neural nets;convolutional neural networks;checkerboard artifacts;linear multirate systems;upsampling layer forward-propagation;cnn;convolution;deconvolution;image resolution;linear systems;signal resolution;steady-state;europe;convolutional neural networks;checkerboard artifacts"],"search_terms":["convolutional","neural","networks","without","checkerboard","artifacts","sugawara","shiota","kiya"],"title":"Convolutional Neural Networks Without Any Checkerboard Artifacts","year":2018,"dataSources":["yiZioZximP7hphDpY","iuBeKSmaES2fHcEE9"]}