This paper presents an approach to segment lesions from brain magnetic resonance images in a fully automatic manner. The proposed idea leverages the strength of classical random walker algorithm and graph cut optimiza...
详细信息
This paper presents an approach to segment lesions from brain magnetic resonance images in a fully automatic manner. The proposed idea leverages the strength of classical random walker algorithm and graph cut optimization technique in a single framework. We demonstrate that formulating a "prior" from a stochastic model can ameliorate the need of manually selected seed selection process in the random walker framework, thus making the algorithm fully automatic in a generic manner. By analytically solving a linear system of equations in the random walk process, initial labelling is an element of [0, 1] of each pixel in the image are computed to obtain the likelihood probability. These probabilities are then used to compute the likelihood for the data fidelity term in the energy function to compute the final segmentation, which is minimized by min cut max flow algorithm. Experimental results show the superiority of the proposed method over the state-of-the-art techniques on publicly available datasets.
Vector quantization (VQ) methods have been used in a wide range of applications for speech, image, and video data. While classic VQ methods often use expectation maximization, in this paper, we investigate the use of ...
详细信息
Vector quantization (VQ) methods have been used in a wide range of applications for speech, image, and video data. While classic VQ methods often use expectation maximization, in this paper, we investigate the use of stochastic optimization employing our recently proposed noise substitution in vector quantization technique. We consider three variants of VQ including additive VQ, residual VQ, and product VQ, and evaluate their quality, complexity and bitrate in speech coding, image compression, approximate nearest neighbor search, and a selection of toy examples. Our experimental results demonstrate the trade-offs in accuracy, complexity, and bitrate such that using our open source implementations and complexity calculator, the best vector quantization method can be chosen for a particular problem.
Deep learning image classifiers are known to be vulnerable to small adversarial perturbations of input images. In this paper, we derive the locally optimal generalized likelihood ratio test based detector for detectin...
详细信息
ISBN:
(纸本)9781728176055
Deep learning image classifiers are known to be vulnerable to small adversarial perturbations of input images. In this paper, we derive the locally optimal generalized likelihood ratio test based detector for detecting stochastic targeted universal adversarial perturbations to a classifier's input. We employ a two-stage process to learn the detector's parameters, which involves unsupervised maximum likelihood estimation followed by supervised training and demonstrates better performance of the detector compared to other detection methods on several popular image classification datasets.
Several of the major issues affecting food productivity are a pest. The timely and precise detection of plant pests is crucial for avoiding the loss of agricultural productivity. Only by detecting the pest at an early...
详细信息
ISBN:
(纸本)9783031243660;9783031243677
Several of the major issues affecting food productivity are a pest. The timely and precise detection of plant pests is crucial for avoiding the loss of agricultural productivity. Only by detecting the pest at an early stage can it be controlled. Due to the cyclical nature of agriculture, pest accumulation and variety might vary from season to season, rendering standard approaches for pest classification and detection ineffective. methods based on machine learning can be utilized to resolve such issues. Deep Learning, which has become extremely popular in imageprocessing, has recently opened up a plethora of new applications for smart agriculture. Optimizers are primarily responsible for the process of strengthening the deep learning model's pest detection capabilities. In order to detect pests on tomato plants, this study compares the performance of a few gradient-based optimizers, including stochastic gradient descent, root means square propagation, adaptive gradient, and adaptive moment estimation, on a proposed deep convolution neural network architecture with augmented data. In comparison to other optimizers, the evaluation findings demonstrate that the Adam optimizer performs better with an accuracy of 93% for pest identification.
Deep steganalyzer combined with neural networks has achieved great success in image classification over recent years. However, it suffers from the following persistent challenges: i) Deep steganalyzer is extremely vul...
详细信息
ISBN:
(数字)9798350374513
ISBN:
(纸本)9798350374520
Deep steganalyzer combined with neural networks has achieved great success in image classification over recent years. However, it suffers from the following persistent challenges: i) Deep steganalyzer is extremely vulnerable and has the risk of being attacked via adversarial steganography when performing the image classification tasks; ii) Pre-processing based methods aiming to remove adversarial perturbations from cover images jeopardize the accuracy performance, as the involved steganographic signal will be wiped off as well. In this context, to defend against adversarial attacks by an adversary, we propose an adversarial steganography detection scheme based on the pre-processing and feature migration. In brief, sub-images are sampled to obtain the dimensionality of the extracted features, which are usually used to expand them while reducing the effect brought by adversarial perturbations. In particular, by computing statistical features together with normalizing the features, our approach can improve the classification accuracy of the samples. Our experimental results show that the proposed approach is capable of detecting adversarial steganographic image with an accuracy gain of up to 35.9% over the state-of-the-art methods.
The accuracy of electric load forecasting is crucial when working on applications in power grid decision-making and operation. Due to the non-linear and stochastic behaviour of customers, the electric load profile is ...
详细信息
ISBN:
(数字)9781665495783
ISBN:
(纸本)9781665495783
The accuracy of electric load forecasting is crucial when working on applications in power grid decision-making and operation. Due to the non-linear and stochastic behaviour of customers, the electric load profile is a complicated signal. In this paper, authors propose machine learning based automated system for electricity load forecasting, taking into consideration various variable factors that have an impact on the future electricity load demand. Three machine learning algorithms are used for evaluation of the proposed framework. The algorithms are evaluated on electricity load data collected from eastern region of Ontario, integrated with the weather and population data of the region. The Light GBM algorithm comparatively performs best with mean absolute error of 0.156. The developed system can be used for more accurate and efficient load forecasting applications.
neural networks have been shown to outperform kernel methods in practice (including neural tangent kernels). Most theoretical explanations of this performance gap focus on learning a complex hypothesis class;in some c...
详细信息
ISBN:
(纸本)9781713845393
neural networks have been shown to outperform kernel methods in practice (including neural tangent kernels). Most theoretical explanations of this performance gap focus on learning a complex hypothesis class;in some cases, it is unclear whether this hypothesis class captures realistic data. In this work, we propose a related, but alternative, explanation for this performance gap in the image classification setting, based on finding a sparse signal in the presence of noise. Specifically, we prove that, for a simple data distribution with sparse signal amidst high-variance noise, a simple convolutional neural network trained using stochastic gradient descent simultaneously learns to threshold out the noise and find the signal. On the other hand, the corresponding neural tangent kernel, with a fixed set of predetermined features, is unable to adapt to the signal in this manner. We supplement our theoretical results by demonstrating this phenomenon empirically: in CIFAR-10 and MNIST images with various backgrounds, as the background noise increases in intensity, a CNN's performance stays relatively robust, whereas its corresponding neural tangent kernel sees a notable drop in performance. We therefore propose the local signal adaptivity (LSA) phenomenon as one explanation for the superiority of neural networks over kernel methods.
Over the past, deep neural networks have proved to be an essential element for developing intelligent solutions. They have achieved remarkable performances at a cost of deeper layers and millions of parameters. Theref...
详细信息
Over the past, deep neural networks have proved to be an essential element for developing intelligent solutions. They have achieved remarkable performances at a cost of deeper layers and millions of parameters. Therefore utilising these networks on limited resource platforms for smart cameras is a challenging task. In this context, models need to be (i) accelerated and (ii) memory efficient without significantly compromising on performance. Numerous works have been done to obtain smaller, faster and accurate models. This paper presents a survey of methods suitable for porting deep neural networks on resource-limited devices, especially for smart cameras. These methods can be roughly divided in two main sections. In the first part, we present compression techniques. These techniques are categorized into: knowledge distillation, pruning, quantization, hashing, reduction of numerical precision and binarization. In the second part, we focus on architecture optimization. We introduce the methods to enhance networks structures as well as neural architecture search techniques. In each of their parts, we describe different methods, and analyse them. Finally, we conclude this paper with a discussion on these methods.
In the last few years, we have witnessed the rise of a series of deep learning methods to generate synthetic images that look extremely realistic. These techniques prove useful in the movie industry and for artistic p...
详细信息
ISBN:
(数字)9781665496209
ISBN:
(纸本)9781665496209
In the last few years, we have witnessed the rise of a series of deep learning methods to generate synthetic images that look extremely realistic. These techniques prove useful in the movie industry and for artistic purposes. However, they also prove dangerous if used to spread fake news or to generate fake online accounts. For this reason, detecting if an image is an actual photograph or has been synthetically generated is becoming an urgent necessity. This paper proposes a detector of synthetic images based on an ensemble of Convolutional neural Networks (CNNs). We consider the problem of detecting images generated with techniques not available at training time. This is a common scenario, given that new image generators are published more and more frequently. To solve this issue, we leverage two main ideas: (i) CNNs should provide "orthogonal" results to better contribute to the ensemble;(ii) the original-image class is better defined than the synthetic-image one, thus it should be better trusted at testing time. Experiments show that pursuing these two ideas improves the detector accuracy on NVIDIA's newly generated StyleGAN3 images, never used in training.
Remote sensing applications are nowadays widely spread in various industrial fields, such as mineral and water exploration, geo-structural mapping, and natural hazards analysis. These applications require that the per...
详细信息
ISBN:
(纸本)9781510657489;9781510657496
Remote sensing applications are nowadays widely spread in various industrial fields, such as mineral and water exploration, geo-structural mapping, and natural hazards analysis. These applications require that the performance of imageprocessing tasks, such as segmentation, object detection, and classification, to be of high accuracy. This can be achieved with relative ease if the given image has high spatial resolution as well as high spectral resolution. However, due to sensor limitations, spatial and spectral resolutions have an inherently inverse relationship and cannot be achieved simultaneously. Hyperspectral images (HSI) have high spectral resolution, but suffer from low spatial resolution, which hinders utilizing them to their full potential. One of the most widely used approaches to enhance spatial resolution is Single image Super Resolution (SISR) techniques. In the recent years, Deep Convolutional neural Networks (DCNNs) have been widely used for HSI enhancement, as they have shown superiority over other traditional methods. Nonetheless, researches still aspire to enhance HSI quality further while overcoming common challenges, such as spectral distortions. Research has shown that properties of natural images can be easily captured using complex numbers. However, this has not been thoroughly investigated from the perspective of HSI SISR. In this paper, we propose a variation of a Complex Valued neural Network (CVNN) architecture for HSI spatial enhancement. The benefits of approaching the problem from a frequency domain perspective will be answered and the proposed network will be compared to its real counterpart and other state-of-the-art approaches. The evaluation and comparison will be recorded qualitatively by visual comparison, and quantitatively using Peak signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), and Spectral Angle Mapper (SAM). The project can be access at: https://***/NourO93/3D_CCVNN_Hyperspectral.
暂无评论