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

Color demosaicking is an ill-posed inverse problem of image restoration. The performance of a color demosaicking algorithm depends on how thoroughly it can exploit domain knowledge to confine the solution space for the underlying true color image. We propose a sparsity-based l 1 minimization technique for color demosaicking that exploits both interband and intra-band sparse representations of natural images. In some of most challenging cases of color demosaicking, the proposed technique outperforms those published in the literature by a significant margin in both PSNR and visual quality.

Color demosaicking with sparse representations

In recent years sparse representation model (SRM) based image deblurring approaches have shown promising image deblurring results. However, since most of the current SRMs don't utilize the spatial correlations between the nonzero sparse coefficients, the SRM-based image deblurring methods often fail to faithfully recover sharp image edges. In this paper, a structured SRM is employed to exploit the local and nonlocal spatial correlation between the sparse codes. The connection between the structured SRM and the low-rank approximation model has also been exploited. An effective image deblurring algorithm using the patch-based structured SRM is then proposed. Experimental results demonstrate the improvements of the proposed deblurring method over current state-of-the-art image deblurring methods.

/pdf/image-deblurring-with-low-rank-approximation-structured-1r81ivb0dg.pdf

Image deblurring with low-rank approximation structured sparse representation

A (v,g)-cage is a (not necessarily unique) smallest v-regular graph of girth g. On such a graph, a nonbinary (2,v)-regular low-density parity-check (LDPC) code can be defined such that the Tanner graph has girth 2g and the code length achieves the minimum possible. In this paper, we focus on two aspects of this class of codes, structural property and code optimization. We find that, in addition to those found previously, many cages can be used to construct structured LDPC codes. We show that all cages with even girth can be structured as protograph-based codes, many of which have block-circulant Tanner graphs. We also find that four cages with odd girth can be structured as protograph-based codes with block-circulant Tanner graphs. For code optimization, we develop an ontology-based approach. All possible inter-connected cycle patterns that lead to low symbol-weight codewords are identified to put together the ontology. By doing so, it becomes handleable to estimate and optimize distance spectrum of equivalent binary image codes. We further analyze some known codes from the Consultative Committee for Space Data Systems recommendation and design several new codes. Numerical results show that these codes have reasonably good minimum bit distance and perform well under iterative decoding.

Nonbinary LDPC Codes on Cages: Structural Property and Code Optimization

Typical image aesthetics assessment (IAA) is modeled for the generic aesthetics perceived by an ``average'' user. However, such generic aesthetics models neglect the fact that users' aesthetic preferences vary significantly depending on their unique preferences. Therefore, it is essential to tackle the issue for personalized IAA (PIAA). Since PIAA is a typical small sample learning (SSL) problem, existing PIAA models are usually built by fine-tuning the well-established generic IAA (GIAA) models, which are regarded as prior knowledge. Nevertheless, this kind of prior knowledge based on ``average aesthetics'' fails to incarnate the aesthetic diversity of different people. In order to learn the shared prior knowledge when different people judge aesthetics, that is, learn how people judge image aesthetics, we propose a PIAA method based on meta-learning with bilevel gradient optimization (BLG-PIAA), which is trained using individual aesthetic data directly and generalizes to unknown users quickly. The proposed approach consists of two phases: 1) meta-training and 2) meta-testing. In meta-training, the aesthetics assessment of each user is regarded as a task, and the training set of each task is divided into two sets: 1) support set and 2) query set. Unlike traditional methods that train a GIAA model based on average aesthetics, we train an aesthetic meta-learner model by bilevel gradient updating from the support set to the query set using many users' PIAA tasks. In meta-testing, the aesthetic meta-learner model is fine-tuned using a small amount of aesthetic data of a target user to obtain the PIAA model. The experimental results show that the proposed method outperforms the state-of-the-art PIAA metrics, and the learned prior model of BLG-PIAA can be quickly adapted to unseen PIAA tasks.

Guangming Shi

Papers

Color demosaicking with sparse representations

Image deblurring with low-rank approximation structured sparse representation

Nonbinary LDPC Codes on Cages: Structural Property and Code Optimization

Personalized Image Aesthetics Assessment via Meta-Learning With Bilevel Gradient Optimization.

Incremental low-rank and sparse decomposition for compressing videos captured by fixed cameras