Search

Segmentation of white matter lesions and deep grey matter structures is an important task in the quantification of magnetic resonance imaging in multiple sclerosis. In this paper we explore segmentation solutions based on convolutional neural networks (CNNs) for providing fast, reliable segmentations of lesions and grey-matter structures in multi-modal MR imaging, and the performance of these methods when applied to out-of-centre data. We trained two state-of-the-art fully convolutional CNN architectures on the 2016 MSSEG training dataset, which was annotated by seven independent human raters: a reference implementation of a 3D Unet, and a more recently proposed 3D-to-2D architecture (DeepSCAN). We then retrained those methods on a larger dataset from a single centre, with and without labels for other brain structures. We quantified changes in performance owing to dataset shift, and changes in performance by adding the additional brain-structure labels. We also compared performance with freely available reference methods. Both fully-convolutional CNN methods substantially outperform other approaches in the literature when trained and evaluated in cross-validation on the MSSEG dataset, showing agreement with human raters in the range of human inter-rater variability. Both architectures showed drops in performance when trained on single-centre data and tested on the MSSEG dataset. When trained with the addition of weak anatomical labels derived from Freesurfer, the performance of the 3D Unet degraded, while the performance of the DeepSCAN net improved. Overall, the DeepSCAN network predicting both lesion and anatomical labels was the best-performing network examined.

Specialised pest and disease control in the agricultural crops industry have been a high-priority issue. Due to great cost-effectiveness and efficient automation, computer vision (CV)–based automatic pest or disease identification techniques are widely utilised in the smart agricultural systems. As rapid development of artificial intelligence, in the field of computer vision–based agricultural pest identification, an increasing number of scholars have begun to move their attentions from traditional machine learning models to deep learning techniques. However, so far, deep learning techniques still have been suffering from many problems such as limited data samples, cost-effectiveness of network structure, and high image quality requirements. These issues greatly limit the potential utilisation of deep-learning techniques into smart agricultural systems. This paper aims at investigating utilization of one new deep-learning model WRN (wide residual networks) into CV-based automatic disease identification problem. We first built up a large-scale agricultural disease images dataset containing over 36,000 pieces of diseases, which includes typical types of disease in tomato, potato, grape, corn and apple. Then, we analysed and evaluated wide residual networks algorithm using the Tesla K80 graphics processor (GPU) in the TensorFlow deep-learning framework. A set of comprehensive experimental protocols have been designed in comparing with GoogLeNet Inception V4 regarding several benchmarks. The experimental results indicate that (1) under WRN architecture, Softmax loss function gives a faster convergence and improved accuracy than GoogLeNet inception V4 network. (2) While WRN shows a good effect for identification of agricultural diseases, its effectiveness has a strong need on the number of training samples of dataset like at least 36 k images in our experiment. (3) The overall performance is better than 800 sheets. The disease identification results show that the WRN model can be applied to the identification of agricultural diseases.

Sleeping problems have become one of the major diseases all over the world. To tackle this issue, the basic tool used by specialists is the Polysomnogram, which is a collection of different signals recorded during sleep. After its recording, the specialists have to score the different signals according to one of the standard guidelines. This process is carried out manually, which can be highly time-consuming and very prone to annotation errors. Therefore, over the years, many approaches have been explored in an attempt to support the specialists in this task. In this paper, an approach based on convolutional neural networks is presented, where an in-depth comparison is performed in order to determine the convenience of using more than one signal simultaneously as input. Additionally, the models were also used as parts of an ensemble model to check whether any useful information can be extracted from signal processing a single signal at a time which the dual-signal model cannot identify. Tests have been performed by using a well-known dataset called expanded sleep-EDF, which is the most commonly used dataset as the benchmark for this problem. The tests were carried out with a leave-one-out cross-validation over the patients, which ensures that there is no possible contamination between training and testing. The resulting proposal is a network smaller than previously published ones, but which overcomes the results of any previous models on the same dataset. The best result shows an accuracy of 92.67\% and a Cohen's Kappa value over 0.84 compared to human experts., Comment: 20 pages, 4 figures, 4 tables

Texture is a fundamental characteristic of many types of images, and texture representation is one of the essential and challenging problems in computer vision and pattern recognition which has attracted extensive research attention over several decades. Since 2000, texture representations based on Bag of Words and on Convolutional Neural Networks have been extensively studied with impressive performance. Given this period of remarkable evolution, this paper aims to present a comprehensive survey of advances in texture representation over the last two decades. More than 250 major publications are cited in this survey covering different aspects of the research, including benchmark datasets and state of the art results. In retrospect of what has been achieved so far, the survey discusses open challenges and directions for future research.

Recent CNNs based salient object detection approaches tend to embed a fully connected Conditional Random Field (CRF) layer to refine the saliency maps from CNNs for post processing. Due to the significant performance enhancement by the CRF layer, in this paper, we propose a more flexible CRF refinement framework by embedding the CRF inference to multiple levels of side outputs from CNNs for multi-scale saliency refinement. A fully convolutional neural networks based on the simple yet effective encoder-decoder architecture with only three scales of side output maps is pre-trained. Then, the CRF layers are embedded to each scale of the side output respectively to complement the defects of each side output maps. Finally, the refined side output maps are fused and refined by another CRF inference for the final saliency map. The proposed multi-scale CRFs model (MCRF) is trained with low computational costs and shows competitive performance over four datasets in comparison with the existing state-of-the-art saliency models.

Devising computational models for detecting abnormalities reflective of diseases from facial structures is a novel and emerging field of research in automatic face analysis. In this paper, we focus on automatic pain intensity estimation from faces. This has a paramount potential diagnosis values in healthcare applications. In this context, we present a novel 3D deep model for dynamic spatiotemporal representation of faces in videos. Using several convolutional layers with diverse temporal depths, our proposed model captures a wide range of spatiotemporal variations in the faces. Moreover, we introduce a cross-architecture knowledge transfer technique for training 3D convolutional neural networks using a pre-trained 2D architecture. This strategy is a practical approach for training 3D models, especially when the size of the database is relatively small. Our extensive experiments and analysis on two benchmarking and publicly available databases, namely the UNBC-McMaster shoulder pain and the BioVid, clearly show that our proposed method consistently outperforms many state-of-the-art methods in automatic pain intensity estimation.

In this work we propose a neural network based image descriptor suitable for image patch matching, which is an important task in many computer vision applications. Our approach is influenced by recent success of deep convolutional neural networks (CNNs) in object detection and classification tasks. We develop a model which maps the raw input patch to a low dimensional feature vector so that the distance between representations is small for similar patches and large otherwise. As a distance metric we utilize L2 norm, i.e. Euclidean distance, which is fast to evaluate and used in most popular hand-crafted descriptors, such as SIFT. According to the results, our approach outperforms state-of-the-art L2-based descriptors and can be considered as a direct replacement of SIFT. In addition, we conducted experiments with batch normalization and histogram equalization as a preprocessing method of the input data. The results confirm that these techniques further improve the performance of the proposed descriptor. Finally, we show promising preliminary results by appending our CNNs with recently proposed spatial transformer networks and provide a visualisation and interpretation of their impact., Comment: The paper was published in ACCV 2016 Workshops proceedings (Workshop on Interpretation and Visualization of Deep Neural Nets)

This paper proposed a novel sequence image representation method called concatenated frame image (CFI), two types of data augmentation methods for CFI, and a framework of CFI-based convolutional neural network (CNN) for visual speech recognition (VSR) task. CFI is a simple, however, it contains spatial-temporal information of a whole image sequence. The proposed method was evaluated with a public database OuluVS2. This is a multi-view audio-visual dataset recorded from 52 subjects. The speaker independent recognition tasks were carried out with various experimental conditions. As the result, the proposed method obtained high recognition accuracy.

Many machine learning softwares are available which help the researchers to accomplish various tasks. These software packages have various conventional algorithms which perform well if the training and test data are independent and identically distributed. However, this might not be the case in the real world. The training data may not be available at one time. In the case of neural networks, the architecture has to be retrained with new data that are made available subsequently. In this paper, we present a novel training algorithm which can avoid complete retraining of any neural network architecture meant for visual pattern recognition. To show the utility of the algorithm, we have investigated the performance of convolutional neural network (CNN) architecture for a face recognition task under transfer learning. The proposed training algorithm may be used for enhancing the utility of machine learning software by providing researchers with an approach that can reduce the training time under transfer learning.

The paper reviews and extends an emerging body of theoretical results on deep learning including the conditions under which it can be exponentially better than shallow learning. A class of deep convolutional networks represent an important special case of these conditions, though weight sharing is not the main reason for their exponential advantage. Implications of a few key theorems are discussed, together with new results, open problems and conjectures., McGovern Institute for Brain Research at MIT. Center for Brains, Minds, and Machines, National Science Foundation (U.S.) (STC award CCF (No. 1231216)), United States. Army Research Office (No. W911NF-15-1-0385)

Deep learning is an important component of big-data analytic tools and intelligent applications, such as, self-driving cars, computer vision, speech recognition, or precision medicine. However, the training process is computationally intensive, and often requires a large amount of time if performed sequentially. Modern parallel computing systems provide the capability to reduce the required training time of deep neural networks. In this paper, we present our parallelization scheme for training convolutional neural networks (CNN) named Controlled Hogwild with Arbitrary Order of Synchronization (CHAOS). Major features of CHAOS include the support for thread and vector parallelism, non-instant updates of weight parameters during back-propagation without a significant delay, and implicit synchronization in arbitrary order. CHAOS is tailored for parallel computing systems that are accelerated with the Intel Xeon Phi. We evaluate our parallelization approach empirically using measurement techniques and performance modeling for various numbers of threads and CNN architectures. Experimental results for the MNIST dataset of handwritten digits using the total number of threads on the Xeon Phi show speedups of up to 103x compared to the execution on one thread of the Xeon Phi, 14x compared to the sequential execution on Intel Xeon E5, and 58x compared to the sequential execution on Intel Core i5., The Journal of Supercomputing, 2017

In recent years, deep learning has not only permeated the computer vision and speech recognition research fields but also fields such as acoustic event detection (AED). One of the aims of AED is to detect and classify non-speech acoustic events occurring in conversation scenes including those produced by both humans and the objects that surround us. In AED, deep learning has enabled modeling of detail-rich features, and among these, high resolution spectrograms have shown a significant advantage over existing predefined features (e.g., Mel-filter bank) that compress and reduce detail. In this paper, we further asses the importance of feature extraction for deep learning-based acoustic event detection. AED, based on spectrogram-input deep neural networks, exploits the fact that sounds have “global” spectral patterns, but sounds also have “local” properties such as being more transient or smoother in the time-frequency domain. These can be exposed by adjusting the time-frequency resolution used to compute the spectrogram, or by using a model that exploits locality leading us to explore two different feature extraction strategies in the context of deep learning: (1) using multiple resolution spectrograms simultaneously and analyzing the overall and event-wise influence to combine the results, and (2) introducing the use of convolutional neural networks (CNN), a state of the art 2D feature extraction model that exploits local structures, with log power spectrogram input for AED. An experimental evaluation shows that the approaches we describe outperform our state-of-the-art deep learning baseline with a noticeable gain in the CNN case and provides insights regarding CNN-based spectrogram characterization for AED.

Double JPEG compression detection has received considerable attention in blind image forensics. However, only few techniques can provide automatic localization. To address this challenge, this paper proposes a double JPEG compression detection algorithm based on a convolutional neural network (CNN). The CNN is designed to classify histograms of discrete cosine transform (DCT) coefficients, which differ between single-compressed areas (tampered areas) and double-compressed areas (untampered areas). The localization result is obtained according to the classification results. Experimental results show that the proposed algorithm performs well in double JPEG compression detection and forgery localization, especially when the first compression quality factor is higher than the second.