RDCGAN: Unsupervised Representation Learning With Regularized Deep Convolutional Generative Adversarial Networks

Mehralian, Mehran; Karasfi, Babak

doi:10.1109/aiar.2018.8769811

Cited by 337 publications

(449 citation statements)

References 6 publications

Supporting

Mentioning

441

Contrasting

Unclassified

Order By: Relevance

“…Because small (negative) value will be induced in the down-sampling model, Leaky ReLU was used in the encoder network (downsampling) to prevent the "dead ReLU" problem. However, for the decoder network (up-sampling), RELU activation function still work well and have the advantages of faster performance and introduce more non-linearity [17], [26]. For the output layer, tanh activation function was implemented to normalized the output in the range of [-1,1] since the range of the brightness/value of each pixel of real image should be within [-1,1].…”

Section: B Pix2pix Gan-based Fd-octmentioning

confidence: 99%

Applying a Pix2Pix Generative Adversarial Network to a Fourier-Domain Optical Coherence Tomography System for Artifact Elimination

2021

View full text Add to dashboard Cite

The presence of artifacts, including conjugate, DC, and auto-correlation artifacts, is a critical limitation of Fourier-domain optical coherence tomography (FD-OCT). Many methods have been proposed to resolve this problem to obtain high-quality images. Furthermore, the development of deep learning has resulted in many prospective advancements in the medical field; image-to-image translation by using generative adversarial networks (GANs) is one such advancement. In this study, we propose applying the Pix2Pix GAN to eliminate artifacts from FD-OCT images. The first experiment results showed that the proposed framework could translate conventional FD-OCT depth profiles into artifact-free FD-OCT depth profiles. In addition, the FD-OCT depth profile and optical distance of translated images matched those of ground truth images. Second experiment verified that the proposed GAN-based FD-OCT can be applied to generate artifact-free FD-OCT image with different parameters of sample refractive index, the front surface of the sample toward the zero-delay position, and the physical thickness of the sample. Third experiment proved that the proposed model could translated the conventional FD-OCT depth profiles with additional Gaussian noises source image into artifacts-free FD-OCT and successfully relieved the noise.INDEX TERMS Artifacts, FD-OCT, image-to-image translation, Pix2Pix GAN the FD-OCT system is more sensitive around the zero-delay line, imaging is performed by positioning the zero-delay line at the region of interest in a sample to obtain double-depth range images. Sanjay et al.[1] used spectral-domain OCT (SD-OCT) to examine patients diagnosed as having glaucoma and reported that 15.2%-36% of scans showed artifacts that may cause difficulty to physicians in the analysis of images. A study identified various types of artifacts that can lead to an incorrect diagnosis [2]. Because artifacts can obscure imaging results and prevent the detection of critical features in a sample structure, full-range FD-OCT is commonly implemented using phase shifting to reconstruct the sample structure and overcome this complex artifact problem. Full-range FD-OCT images are obtained by recording several interferograms with different phase relations. Jiewen et al. [3] proposed a five-frame variable phase-shifting (FVP) method to reduce the effects of polychromatic errors. Compared with the traditional five-

show abstract

Section: B Pix2pix Gan-based Fd-octmentioning

confidence: 99%

Applying a Pix2Pix Generative Adversarial Network to a Fourier-Domain Optical Coherence Tomography System for Artifact Elimination

2021

View full text Add to dashboard Cite

show abstract

“…Generative Adversarial Networks with deep convolutional neural network . [16] is the extension of GAN introduced first. DCGAN can maintain stability in the training process and create high resolution images.…”

Section: Dcganmentioning

confidence: 99%

“…During training process, the random noise is given as the input to the generator and the generator with multi deconvolutional neural network produces the sample images looks like real images. Discriminator tries to differentiate generated images and training set of images [16]. DCGAN uses Batch Norm [17] for normalization of extracted data, and Leaky ReLU [18] for preventing dead gradients.…”

Section: Dcganmentioning

confidence: 99%

Review of Medical Image Synthesis using GAN Techniques

Devi

Suganthi

2021

ITM Web Conf.

View full text Add to dashboard Cite

Generative Adversarial Networks (GANs) is one of the vital efficient methods for generating a massive, high-quality artificial picture. For diagnosing particular diseases in a medical image, a general problem is that it is expensive, usage of high radiation dosage, and time-consuming to collect data. Hence GAN is a deep learning method that has been developed for the image to image translation, i.e. from low-resolution to highresolution image, for example generating Magnetic resonance image (MRI) from computed tomography image (CT) and 7T from 3T MRI which can be used to obtain multimodal datasets from single modality. In this review paper, different GAN architectures were discussed for medical image analysis.

show abstract

“…Compared with the histogram of oriented gradient (HOG) [4], scale-invariant feature transform (SIFT) [5], and so on, the convolutional neural network (CNN) has been broadly studied. Many different CNN architectures and methods have been proposed, for example, AlexNet, VGGNet, GAN, ResNet, and SENet [6][7][8][9][10].…”

Section: Introductionmentioning

confidence: 99%

“…The pooling layer can reduce the computation of CNN training and highlight some important features in the image. Some scholars have also designed the up-sampling method based on the characteristics of the pooling layer, which was used to explain CNN [17], such as a generative adversarial network (GAN) [18], and so on [8], [19]. In general, the pooling operations are maximum pooling and average pooling [20].…”

Section: Introductionmentioning

confidence: 99%

T–S Fuzzy Model Based Multi-Branch Deep Network Architecture

Wang

et al. 2020

IEEE Access

View full text Add to dashboard Cite

In the traditional CNN design, the hyperparameters, such as the size of the convolutional kernel and stride, are difficult to determine. In this paper, a new convolutional network architecture, named multi-branch fuzzy architecture network (MBFAN), was proposed for this problem. In MBFAN, some branches with a certain convolutional neural network architecture are connected in parallel. In each branch, a different-sized convolutional kernel is applied. By data training and normalization, a weight is given to each branch. By these weights, the important features in the final output are strengthened. By normalization, the branches were interconnected together, making the training process more efficient. Due to overfitting, with the increase of branches, the MBFAN accuracy increases, and then decreases. The number of branches is optimized when the MBFAN accuracy is highest. On the other hand, the location of the convolutional kernel center in an image has a great influence on the convolutional results. This is also discussed in MBFAN. For the experiments, the proposed MBFAN was adopted and tested in a simple convolutional network and a VGG16 network. INDEX TERMS deep learning, convolutional neural network, multi-branch, fuzzy.

show abstract

RDCGAN: Unsupervised Representation Learning With Regularized Deep Convolutional Generative Adversarial Networks

Cited by 337 publications

References 6 publications

Applying a Pix2Pix Generative Adversarial Network to a Fourier-Domain Optical Coherence Tomography System for Artifact Elimination

Applying a Pix2Pix Generative Adversarial Network to a Fourier-Domain Optical Coherence Tomography System for Artifact Elimination

Review of Medical Image Synthesis using GAN Techniques

T–S Fuzzy Model Based Multi-Branch Deep Network Architecture

Contact Info

Product

Resources

About