We propose a deep convolutional neural network architecture codenamed Inception that achieves the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC14). The main hallmark of this architecture is the improved utilization of the computing resources inside the network. By a carefully crafted design, we increased the depth and width of the network while keeping the computational budget constant. To optimize quality, the architectural decisions were based on the Hebbian principle and the intuition of multi-scale processing. One particular incarnation used in our submission for ILSVRC14 is called GoogLeNet, a 22 layers deep network, the quality of which is assessed in the context of classification and detection.

/pdf/going-deeper-with-convolutions-1yobw2o2ds.pdf

Going deeper with convolutions

[신간의 별자리x] 우리/미술, 그리고 ‘슬픔의 박물관’

Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Pattern Recognition and Machine Learning

I have developed "tennis elbow" from lugging this book around the past four weeks, but it is worth the pain, the effort, and the aspirin. It is also worth the (relatively speaking) bargain price. Including appendixes, this book contains 894 pages of text. The entire panorama of the neural sciences is surveyed and examined, and it is comprehensive in its scope, from genomes to social behaviors. The editors explicitly state that the book is designed as "an introductory text for students of biology, behavior, and medicine," but it is hard to imagine any audience, interested in any fragment of neuroscience at any level of sophistication, that would not enjoy this book. The editors have done a masterful job of weaving together the biologic, the behavioral, and the clinical sciences into a single tapestry in which everyone from the molecular biologist to the practicing psychiatrist can find and appreciate his or

Principles of Neural Science

Despite the breakthroughs in accuracy and speed of single image super-resolution using faster and deeper convolutional neural networks, one central problem remains largely unsolved: how do we recover the finer texture details when we super-resolve at large upscaling factors? The behavior of optimization-based super-resolution methods is principally driven by the choice of the objective function. Recent work has largely focused on minimizing the mean squared reconstruction error. The resulting estimates have high peak signal-to-noise ratios, but they are often lacking high-frequency details and are perceptually unsatisfying in the sense that they fail to match the fidelity expected at the higher resolution. In this paper, we present SRGAN, a generative adversarial network (GAN) for image super-resolution (SR). To our knowledge, it is the first framework capable of inferring photo-realistic natural images for 4x upscaling factors. To achieve this, we propose a perceptual loss function which consists of an adversarial loss and a content loss. The adversarial loss pushes our solution to the natural image manifold using a discriminator network that is trained to differentiate between the super-resolved images and original photo-realistic images. In addition, we use a content loss motivated by perceptual similarity instead of similarity in pixel space. Our deep residual network is able to recover photo-realistic textures from heavily downsampled images on public benchmarks. An extensive mean-opinion-score (MOS) test shows hugely significant gains in perceptual quality using SRGAN. The MOS scores obtained with SRGAN are closer to those of the original high-resolution images than to those obtained with any state-of-the-art method.

/pdf/photo-realistic-single-image-super-resolution-using-a-2y0438rlmd.pdf

Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network

In this brief, the multiplier-free implementation of the constant vector multiplication is reexamined. A novel improved signed digit representation technique is proposed to overcome the two main drawbacks of the current multiplier-free techniques: 1) computational redundancy and 2) circuit irregularity. The fundamental difference between the proposed technique and the existing multiplier-free techniques is a novel optimization framework based on vector decomposition. The constant vector is decomposed into two terms: a “public” vector and a “private” matrix which consist of the public operations shared by all of the entries and the private operations of each individual entry, respectively. In this way, the overall data flow can be divided into two regular steps: multiplied by the “public” vector first and then by the “private” matrix. The computational complexity reduction task is then achieved by minimizing the length of the “public” vector and the number of operations in the “private” matrix. Experimental results demonstrate that the proposed technique outperforms the existing multiplier-free techniques in fewer operations and more regular circuit structure.

An Improved Signed Digit Representation Approach for Constant Vector Multiplication

Canonical multi-input–multi-output (MIMO) imaging methods suffer from limited resolution, poor robustness against noise and high computational complexity, especially when the array aperture is limited (affecting angular resolution) and the number of snapshots is limited (affecting Doppler resolution). In this study, the authors discuss a new range-angle-Doppler monostatic MIMO imaging method through adaptive estimation of the generalised Cauchy prior distribution (GCD). The superiority of GCD-based model over existing l p -norm-based model (which actually assumes the prior as general Gaussian distribution) is theoretically verified through the issue of signal compressibility. The authors adapt a reweighted l2-norm iterative algorithm to solve the model. In our model the authors do not pre-define the scaling factor of the prior distribution and, during the iteration, the authors use a novel ‘quantile-of-OS’ method to adaptively estimate the scaling parameter of the prior distribution, enhancing the robustness of the method. Simulation results verify the image quality and speed advantages of the proposed method.

https://ietresearch.onlinelibrary.wiley.com/doi/pdf/10.1049/iet-rsn.2011.0398

Robust compressive multi-input–multi-output imaging

The invention discloses a method for obtaining the depth of a structured light dynamic scene on the basis of random templates. The method mainly solves the problems that the prior art is low in depth spatial resolution and depth precision. The method comprises the implementation steps of designing the random templates P, projecting the random templates P into a three-dimensional scene through a projector T, and recording deformed images U which are modulated through the scene by using a video camera C; averaging the random templates P and deformed image U sliding windows to obtain stripe-liked images Q and deformed stripe images B; solving the phase positions of pixels in the stripe-liked images Q and in the deformed stripe images B by using a Gabor filter; obtaining a rough matching result of the pixels in the deformed stripe images B and the pixels in the stripe-liked images Q through a phase relation; carrying out compensation on the rough matching result to obtain an ultimate accurate matching result; solving the depth of the corresponding scene by utilizing coordinates of matching points and a line and plane intersection geometrical relationship. The method for obtaining the depth of the structured light dynamic scene on the basis of the random template has the advantages of being high in depth spatial resolution and low in calculation complexity, and is capable of being applied to accurate three-dimensional reconstruction of the dynamic scene.

Method for obtaining depth of structured light dynamic scene on basis of random templates

The recently proposed compressive sensing (CS) theory provides a new solution for multiple description coding (MDC) with fine granularity, by treating each random CS measurement as a description. The performance of CS-based MDC (CS-MDC) depends on the efficacy of the CS recovery algorithm. Existing CS recovery algorithms recover the signal in a fixed space (e.g., Wavelet, DCT, and gradient spaces) for the entire duration of the signal, even though a typical multimedia signal exhibits sparsity in time/space variant spaces. To rectify this problem and develop a better CS recovery algorithm for CSMDC, we propose a learning-based framework to conduct the CS recovery in locally adaptive spaces, and carry out a case study on image MDC. A set of prior image models are learned offline from a training set to facilitate the CS recovery in local adaptive bases. Experiments show that the learning-based CS recovery algorithm can significantly improve the performance of the previous CS-MDC technique in both PSNR and visual quality.

Learning-based recovery of compressive sensing with application in multiple description coding

Zero-shot learning (ZSL), which focuses on transferring the knowledge from the seen classes to unseen ones, has attracted more and more attention in the computer vision community. Exploring the relationships among the spaces of visual representation, semantic description and label information is a key to the success of ZSL. In this paper, we propose a novel approach by using a two-layer Stacked AutoEncoder (StAE) with manifold regularizations to construct the tight relations of different spaces, where the first-layer encoder aims to project a visual feature vector into the semantic space, and the second-layer encoder connects the semantic description of a sample with its label directly. Meanwhile, the decoders seek to reconstruct the visual representation from label information and semantic description successively. Besides, two manifold regularizers are integrated in the stacked autoencoder, which captures the manifold structures residing in the different spaces effectively. Compared with the previous related works, the proposed approach is a more general framework and has stronger transfer ability from seen classes to unseen classes. Extensive experiments on the benchmark datasets clearly demonstrate that our StAE performs significantly better than the state-of-the-arts.

Guangming Shi

Papers

An Improved Signed Digit Representation Approach for Constant Vector Multiplication

Robust compressive multi-input–multi-output imaging

Method for obtaining depth of structured light dynamic scene on basis of random templates

Learning-based recovery of compressive sensing with application in multiple description coding

Zero-Shot Learning Using Stacked Autoencoder with Manifold Regularizations