Classification of massive noisy image using auto-encoders and convolutional neural network

Roy, Sudipta; Ahmed, Mahtab; Akhand, M. A. H.

doi:10.1109/icitech.2017.8079976

Cited by 11 publications

(15 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The parameters of the network, as well as the kernel, get updated during the training process until the desired accuracy is achieved. The detail description of CNN training is also available in the previous studies [2], [42]. 228 | P a g e www.ijacsa.thesai.org Fig.…”

Section: ) Convolutional Neural Network (Cnn): Cnnmentioning

confidence: 99%

“…The DAE is trained with regular noise only and can"t reconstruct native images which are corrupted with massive level of noises, such as if the percentage of noise in the images are around 50%. Roy et al [42] showed that cascaded architectures of DAEs, where each of the DAEs is trained with 20% noisy images, can reconstruct images of good quality even if the noise level present in the images is 50%. Following this idea, a cascaded DAE-DAE is arranged as the image denoiser in the third structure DAE-DAE-CNN.…”

Section: B Proposed Robust System Combining Different Structures Witmentioning

confidence: 99%

“…During backpropagation, this model learns to filter out the underlying noise from the input image and reconstruct a noiseless one. The detailed description, as well as training of DAE, is available in the previous studies [42]. [19] is a variant of neural network popular for object detection and segmentation task.…”

Section: Z= (W´y+b´)mentioning

confidence: 99%

“…Since CNN is trained with native images without noise it is well for noiseless image classification. On the other hand, DAE-CNN and DAE-DAE-CNN structures perform well for low-level and high-level noisy cases, respectively [42]. In DAE-CNN, DAE first removes noises from noisy input images and then CNN is fed with these restored images for classification purpose.…”

Section: Introductionmentioning

confidence: 99%

“…Additionally, they embedded a whitening layer in front of each and every convolutional layer to enclose the input feature maps. Most recently, Roy et al [42] applied convolutional denoising autoencoder (CDAE) followed by a DAE and arranged them in a cascaded manner, rather than in a stacked way to deal with data subject to massive noise. They showed that if two AE based models are individually trained to denoise images subject to regular level of noise, the cascaded architecture of them can show a great performance in case of denoising massive noisy images.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

A Robust System for Noisy Image Classification Combining Denoising Autoencoder and Convolutional Neural Network

Roy¹,

Hossain²,

Akhand³

et al. 2018

ijacsa

View full text Add to dashboard Cite

Abstract-Image classification, a complex perceptual task with many real life important applications, faces a major challenge in presence of noise. Noise degrades the performance of the classifiers and makes them less suitable in real life scenarios. To solve this issue, several researches have been conducted utilizing denoising autoencoder (DAE) to restore original images from noisy images and then Convolutional Neural Network (CNN) is used for classification. The existing models perform well only when the noise level present in the training set and test set are same or differs only a little. To fit a model in real life applications, it should be independent to level of noise. The aim of this study is to develop a robust image classification system which performs well at regular to massive noise levels. The proposed method first trains a DAE with low-level noise-injected images and a CNN with noiseless native images independently. Then it arranges these two trained models in three different combinational structures: CNN, DAE-CNN, and DAE-DAE-CNN to classify images corrupted with zero, regular and massive noises, accordingly. Final system outcome is chosen by applying the winner-takes-all combination on individual outcomes of the three structures. Although proposed system consists of three DAEs and three CNNs in different structure layers, the DAEs and CNNs are the copy of same DAE and CNN trained initially which makes it computationally efficient as well. In DAE-DAE-CNN, two identical DAEs are arranged in a cascaded structure to make the structure well suited for classifying massive noisy data while the DAE is trained with low noisy image data. The proposed method is tested with MNIST handwritten numeral dataset with different noise levels. Experimental results revealed the effectiveness of the proposed method showing better results than individual structures as well as the other related methods.

show abstract

Section: ) Convolutional Neural Network (Cnn): Cnnmentioning

confidence: 99%

Section: B Proposed Robust System Combining Different Structures Witmentioning

confidence: 99%

Section: Z= (W´y+b´)mentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

A Robust System for Noisy Image Classification Combining Denoising Autoencoder and Convolutional Neural Network

Roy¹,

Hossain²,

Akhand³

et al. 2018

ijacsa

View full text Add to dashboard Cite

show abstract

A survey of multimodal hybrid deep learning for computer vision: Architectures, applications, trends, and challenges

Bayoudh

2024

Information Fusion

View full text Add to dashboard Cite

Neural networks can learn to utilize correlated auxiliary noise

Ahmadzadegan

Simidzija

et al. 2021

Sci Rep

View full text Add to dashboard Cite

We demonstrate that neural networks that process noisy data can learn to exploit, when available, access to auxiliary noise that is correlated with the noise on the data. In effect, the network learns to use the correlated auxiliary noise as an approximate key to decipher its noisy input data. An example of naturally occurring correlated auxiliary noise is the noise due to decoherence. Our results could, therefore, also be of interest, for example, for machine-learned quantum error correction.

show abstract

Classification of massive noisy image using auto-encoders and convolutional neural network

Cited by 11 publications

References 5 publications

A Robust System for Noisy Image Classification Combining Denoising Autoencoder and Convolutional Neural Network

A Robust System for Noisy Image Classification Combining Denoising Autoencoder and Convolutional Neural Network

A survey of multimodal hybrid deep learning for computer vision: Architectures, applications, trends, and challenges

Neural networks can learn to utilize correlated auxiliary noise

Contact Info

Product

Resources

About