Deep Learning For Image Classification Research Articles

A multiclass TrAdaBoost transfer learning algorithm for the classification of mobile lidar data

A major challenge in the application of state-of-the-art deep learning methods to the classification of mobile lidar data is the lack of sufficient training samples for different object categories. The transfer learning technique based on pre-trained networks, which is widely used in deep learning for image classification, is not directly applicable to point clouds, because pre-trained networks trained by a large number of samples from multiple sources are not available. To solve this problem, we design a framework incorporating a state-of-the-art deep learning network, i.e. VoxNet, and propose an extended Multiclass TrAdaBoost algorithm, which can be trained with complementary training samples from other source datasets to improve the classification accuracy in the target domain. In this framework, we first train the VoxNet model with the combined dataset and extract the feature vectors from the fully connected layer, and then use these to train the Multiclass TrAdaBoost. Experimental results show that the proposed method achieves both improvement in the overall accuracy and a more balanced performance in each category.

Deep Learning for Fingerprint Localization in Indoor and Outdoor Environments

Wi-Fi and magnetic field fingerprinting-based localization have gained increased attention owing to their satisfactory accuracy and global availability. The common signal-based fingerprint localization deteriorates due to well-known signal fluctuations. In this paper, we proposed a Wi-Fi and magnetic field-based localization system based on deep learning. Owing to the low discernibility of magnetic field strength (MFS) in large areas, the unsupervised learning density peak clustering algorithm based on the comparison distance (CDPC) algorithm is first used to pick up several center points of MFS as the geotagged features to assist localization. Considering the state-of-the-art application of deep learning in image classification, we design a location fingerprint image using Wi-Fi and magnetic field fingerprints for localization. Localization is casted in a proposed deep residual network (Resnet) that is capable of learning key features from a massive fingerprint image database. To further enhance localization accuracy, by leveraging the prior information of the pre-trained Resnet coarse localizer, an MLP-based transfer learning fine localizer is introduced to fine-tune the coarse localizer. Additionally, we dynamically adjusted the learning rate (LR) and adopted several data enhancement methods to increase the robustness of our localization system. Experimental results show that the proposed system leads to satisfactory localization performance both in indoor and outdoor environments.

Evaluating Deep Learning for Image Classification in Adversarial Environment

Evaluating Deep Learning for Image Classification in Adversarial Environment

A Novel Outdoor Positioning Technique Using LTE Network Fingerprints.

In recent years, wireless-based fingerprint positioning has attracted increasing research attention owing to its position-related features and applications in the Internet of Things (IoT). In this paper, by leveraging long-term evolution (LTE) signals, a novel deep-learning-based fingerprint positioning approach is proposed to solve the problem of outdoor positioning. Considering the outstanding performance of deep learning in image classification, LTE signal measurements are converted into location grayscale images to form a fingerprint database. In order to deal with the instability of LTE signals, prevent the gradient dispersion problem, and increase the robustness of the proposed deep neural network (DNN), the following methods are adopted: First, cross-entropy is used as the loss function of the DNN. Second, the learning rate of the proposed DNN is dynamically adjusted. Third, this paper adopted several data enhancement techniques. To find the best positioning fingerprint and method, three types of fingerprint and five positioning models are compared. Finally, by using a deep residual network (Resnet) and transfer learning, a hierarchical structure training method is proposed. The proposed Resnet is used to train with the united fingerprint image database to obtain a positioning model called a coarse localizer. By using the prior knowledge of the pretrained Resnet, feed-forward neural network (FFNN)-based transfer learning is used to train with the united fingerprint database to obtain a better positioning model, called a fine localizer. The experimental results convincingly show that the proposed DNN can automatically learn the location features of LTE signals and achieve satisfactory positioning accuracy in outdoor environments.

A weakly supervised CNN model for spatial localization of human activities in unconstraint environment

Human action localization in a given video sequences refers to the spatial and temporal information of the specified action. Similar to its recognition, action localization also plays very important roles in security, disease diagnosis and geographical systems. The necessity of its localization can help in tracking, detection and prediction issues of the concerned event. The main issue is noticed, while it has to process long, untrimmed and highly occluded videos in uncontrolled conditions as it requires expensive as well as laborious tasks of retrieving annotation for every action. Motivated from the recent state of the art in deep learning for image classification, we presented a weakly supervised action localization model based on deep neural network. The proposed model is useful in case of large amount of dealing with large amount of data as developing a big network consumes more computational resources and many times it raises overfitting issues. We utilized the effectiveness of Inception V3 model (GoogLeNet) framework which uses TensorFlow at backend and Batch normalization along with the convolution layers. Batch normalization efficiently removes covariant shifts problem between the network layers. The approach developed in this work is validated on UCF50 and UCF sports action benchmark datasets. The proposed model gives satisfactory results as observed from the two data-samples (UCF50 and UCF sports); it can perform better on long untrimmed video sequences captured from unconstraint environment. The important application of this work can be found in very sensitive tasks, like hidden objects auto-localization and detecting enemy position under camera surveillance.

Limits on transfer learning from photographic image data to X-ray threat detection.

X-ray imaging is a crucial and ubiquitous tool for detecting threats to transport security, but interpretation of the images presents a logistical bottleneck. Recent advances in Deep Learning image classification offer hope of improving throughput through automation. However, Deep Learning methods require large quantities of labelled training data. While photographic data is cheap and plentiful, comparable training sets are seldom available for the X-ray domain. To determine whether and to what extent it is feasible to exploit the availability of photo data to supplement the training of X-ray threat detectors. A new dataset was collected, consisting of 1901 matched pairs of photo & X-ray images of 501 common objects. Of these, 258 pairs were of 69 objects considered threats in the context of aviation. This data was used to test a variety of transfer learning approaches. A simple model of threat cue availability was developed to understand the limits of this transferability. Appearance features learned from photos provide a useful basis for training classifiers. Some transfer from the photo to the X-ray domain is possible as ∼40% of danger cues are shared between the modalities, but the effectiveness of this transfer is limited since ∼60% of cues are not. Transfer learning is beneficial when X-ray data is very scarce-of the order of tens of training images in our experiments-but provides no significant benefit when hundreds or thousands of X-ray images are available.

COVID-19 Detection Through Transfer Learning Using Multimodal Imaging Data.

Detecting COVID-19 early may help in devising an appropriate treatment plan and disease containment decisions. In this study, we demonstrate how transfer learning from deep learning models can be used to perform COVID-19 detection using images from three most commonly used medical imaging modes X-Ray, Ultrasound, and CT scan. The aim is to provide over-stressed medical professionals a second pair of eyes through intelligent deep learning image classification models. We identify a suitable Convolutional Neural Network (CNN) model through initial comparative study of several popular CNN models. We then optimize the selected VGG19 model for the image modalities to show how the models can be used for the highly scarce and challenging COVID-19 datasets. We highlight the challenges (including dataset size and quality) in utilizing current publicly available COVID-19 datasets for developing useful deep learning models and how it adversely impacts the trainability of complex models. We also propose an image pre-processing stage to create a trustworthy image dataset for developing and testing the deep learning models. The new approach is aimed to reduce unwanted noise from the images so that deep learning models can focus on detecting diseases with specific features from them. Our results indicate that Ultrasound images provide superior detection accuracy compared to X-Ray and CT scans. The experimental results highlight that with limited data, most of the deeper networks struggle to train well and provides less consistency over the three imaging modes we are using. The selected VGG19 model, which is then extensively tuned with appropriate parameters, performs in considerable levels of COVID-19 detection against pneumonia or normal for all three lung image modes with the precision of up to 86% for X-Ray, 100% for Ultrasound and 84% for CT scans.

#Vape: Measuring E-Cigarette Influence on Instagram With Deep Learning and Text Analysis.

E-cigarette use is increasing dramatically among adolescents as social media marketing portrays “vaping” products as healthier alternatives to conventional cigarettes. In September 2018, the Food and Drug Administration (FDA) launched an anti-vaping campaign, in U.S. high schools, on social media and other platforms, emphasizing “The Real Cost” of e-cigarettes. Using a novel deep learning approach, we assessed changes in vaping-related content on Instagram from 2017 to 2019 and drew an inference about the initial impact of the FDA’s Real Cost campaign on Instagram. We collected 245,894 Instagram posts that used vaping-related hashtags (e.g., #vape, #ejuice) in four samples from 2017 to 2019. We compared the “like” count from these posts before and after the FDA campaign. We used deep learning image classification to analyze 49,655 Instagram image posts, separating images of men, women, and vaping devices. We also conducted text analysis and topic modeling to detect the common words and themes in the posted captions. Since September 2018, the FDA-sponsored hashtag #TheRealCost has been used about 50 times per month on Instagram, whereas vaping-related hashtags we tracked were used up to 10,000 times more often. Comparing the pre-intervention (2017, 2018) and post-intervention (2019) samples of vaping-related Instagram posts, we found a three-fold increase in the median “like” count (10 vs. 28) and a 6-fold increase in the proportion of posts with more than 100 likes (2 vs. 15%). Over 70% of Instagram vaping images featured e-juices and devices, with a growing number of images depicting a “pod,” the type of discrete vaping device that delivers high concentration of nicotine and is favored by novice e-cigarette users. In addition, the Instagram analytics data shared by the vaping influencers we interviewed showed underage Instagram users among their followers.

Designing a Structured System for Mammography Reporting

Background: Breast cancer is among the top causes of cancer-related mortality among females in both developed and developing countries. Screening for breast cancer plays a crucial role in the prevention of disease burden. Among various imaging modalities, mammography is currently considered the first step for the detection of breast lesions. The mammography data of breast cancer patients encompass a wide array of texts related to specific visual findings. Accordingly, unorganized mammography documents usually impede physicians’ efforts to fully review the previous information in follow-up visits; hence, this not only affects clinical judgment and treatment planning adversely, but also adds to the financial burdens and workload imposed on the healthcare system. This further highlights the great potential of structured mammography reporting in the management of breast cancer patients. Methods: In the first step, an expert panel consisting of two attending radiologists and a health informatician (all affiliated to the Shiraz University of Medical Sciences, Shiraz, Iran) reviewed related guidelines to determine the appropriate items, ontologies, and standard formats for data entry. These included radiology textbooks and the systematized nomenclature of medicine (SNOMED). In addition, 100 mammography reports were completely examined for data extraction and the creation of a template report. In the next step, the structure of the data registry was discussed among a group of radiologists and breast cancer surgeons during several sessions using Delphi technique. Results: Overall, 119 fields were selected for data entry. Our survey showed that all of the contributing physicians believed that a structured reporting system for mammography can help standardize and reduce reporting time and errors. Following the completion of the template, a user interface was developed by the expert panel for integration in the cloud software workflow, which will be deployed and assessed in the next phase of the project. Conclusion: Structured mammography reporting helps radiologists and surgeons to efficiently and confidently track the management course of their patients. In addition, with the advancement of artificial intelligence, especially deep learning for image classification, clean and labeled image databases can be used for designing computer-assisted decision support systems without significant data preprocessing.

An Application of Deep Learning Image Classification on Landslide Automated Detection with FORMOSAT-2 Satellite Imagery

An Application of Deep Learning Image Classification on Landslide Automated Detection with FORMOSAT-2 Satellite Imagery

Mosquito Larvae Detection using Deep Learning

Dengue cases has become endemic in Malaysia. The cost of operation to exterminate mosquito habitats are also high. To do effective operation, information from community are crucial. But, without knowing the characteristic of Aedes larvae it is hard to recognize the larvae without guide from the expert. The use of deep learning in image classification and recognition is crucial to tackle this problem. The purpose of this project is to conduct a study of characteristics of Aedes larvae and determine the best convolutional neural network model in classifying the mosquito larvae. 3 performance evaluation vector which is accuracy, log-loss and AUC-ROC will be used to measure the model’s individual performance. Then performance category which consist of Accuracy Score, Loss Score, File Size Score and Training Time Score will be used to evaluate which model is the best to be implemented into web application or mobile application. From the score collected for each model, ResNet50 has proved to be the best model in classifying the mosquito larvae species.

OA06.01 Comparison Between Radiomics-Based Machine Learning and Deep Learning Image Classification for Sub-Cm Lung Nodules

OA06.01 Comparison Between Radiomics-Based Machine Learning and Deep Learning Image Classification for Sub-Cm Lung Nodules

Preprocessing Medical Images for Classification using Deep Learning Techniques

Recently, the demand for computer vision techniques is continuously rising because of the development of techniques in decision making pertaining to health sector. Image processing is a subset of computer vision which makes use of algorithms to perform vision emulation to recognize objects. In this study a novel convolutional neural network is configured based on deep learning to classifying Chest x-ray images into five major classes. It addresses an issue of insufficiency in medical images for employing deep learning for image classification. A new augmentation technique superimposing of images helps to generate more new samples from the available images using label-preserving transformations. Data augmentation technique can generate new sample data from the original data using various transforming strategies. Therefore the data augmentation technique helps in accumulating enough data for processing to obtain better performance. The main objective of superimposing of two images is to minimize redundancy and uncertainty in the output image. Therefore the superimposing carried out with original image and a set of various augmented image to obtain better accuracy. Later results of various superimposing techniques are compared and evaluated to demonstrate the better techniques. It is concluded that the proposed techniques can obtain better performance in medical image classification problem.

Towards Fast and Direct Memory Read-out by Multi-beam Scanning Electron Microscopy and Deep Learning Image Classification

Journal Article Towards Fast and Direct Memory Read-out by Multi-beam Scanning Electron Microscopy and Deep Learning Image Classification Get access Kyle Crosby, Kyle Crosby Carl Zeiss Microscopy LLC, Business Development mSEM, One Zeiss Drive, Thornwood, USA Search for other works by this author on: Oxford Academic Google Scholar Tomasz Garbowski, Tomasz Garbowski Carl Zeiss Microscopy GmbH, Business Development mSEM, Oberkochen, Germany Search for other works by this author on: Oxford Academic Google Scholar Stephan Nickell Stephan Nickell Carl Zeiss Microscopy GmbH, Business Development mSEM, Oberkochen, Germany Corresponding author: Stephan.Nickell@Zeiss.com Search for other works by this author on: Oxford Academic Google Scholar Microscopy and Microanalysis, Volume 25, Issue S2, 1 August 2019, Pages 192–193, https://doi.org/10.1017/S1431927619001697 Published: 01 August 2019

Multi-criteria active deep learning for image classification

As a robust and heuristic technique in machine learning, active learning has been established as an effective method for addressing large volumes of unlabeled data; it interactively queries users (or certain information sources) to obtain desired outputs at new data points. With regard to deep learning techniques (e.g., CNN) and their applications (e.g., image classification), labeling work is of great significance as training processes for obtaining parameters in neural networks which requires abundant labeled samples. Although a few active learning algorithms have been proposed for devising certain straightforward sampling strategies (e.g., density, similarity, uncertainty, and label-based measure) for deep learning algorithms, these employ onefold sampling strategies and do not consider the relationship among multiple sampling strategies.To this end, we devised a novel solution “multi-criteria active leep learning”(MCADL) to learn an active learning strategy for deep neural networks in image classification. Our sample selection strategy selects informative samples by considering multiple criteria simultaneously (i.e., density, similarity, uncertainty, and label-based measure). Moreover, our approach is capable of adjusting weights adaptively to fuse the results from multiple criteria effectively by exploring the utilities of the criteria at different training stages. Through extensive experiments on two popular image datasets (i.e., MNIST and CIFAR-10), we demonstrate that our proposed method consistently outperforms highly competitive active learning approaches; thereby, it can be verified that our multi-criteria active learning proposal is rational and our solution is effective.

SS-HCNN: Semi-Supervised Hierarchical Convolutional Neural Network for Image Classification.

The availability of large-scale annotated data and uneven separability of different data categories become two major impediments of deep learning for image classification. In this paper, we present a Semi-Supervised Hierarchical Convolutional Neural Network (SS-HCNN) to address these two challenges. A large-scale unsupervised maximum margin clustering technique is designed, which splits images into a number of hierarchical clusters iteratively to learn cluster-level CNNs at parent nodes and category-level CNNs at leaf nodes. The splitting uses the similarity of CNN features to group visually similar images into the same cluster, which relieves the uneven data separability constraint. With the hierarchical cluster-level CNNs capturing certain high-level image category information, the category-level CNNs can be trained with a small amount of labelled images, and this relieves the data annotation constraint. A novel cluster splitting criterion is also designed which automatically terminates the image clustering in the tree hierarchy. The proposed SS-HCNN has been evaluated on the CIFAR-100 and ImageNet classification datasets. Experiments show that the SS-HCNN trained using a portion of labelled training images can achieve comparable performance with other fully trained CNNs using all labelled images. Additionally, the SS-HCNN trained using all labelled images clearly outperforms other fully trained CNNs.

Contrast-enhanced computed tomography image assessment of cervical lymph node metastasis in patients with oral cancer by using a deep learning system of artificial intelligence

Although the deep learning system has been applied to interpretation of medical images, its application to the diagnosis of cervical lymph nodes in patients with oral cancer has not yet been reported. The purpose of this study was to evaluate the performance of deep learning image classification for diagnosis of lymph node metastasis. The imaging data used for evaluation consisted of computed tomography (CT) images of 127 histologically proven positive cervical lymph nodes and 314 histologically proven negative lymph nodes from 45 patients with oral squamous cell carcinoma. The performance of a deep learning image classification system for the diagnosis of lymph node metastasis on CT images was compared with the diagnostic interpretations of 2 experienced radiologists by using the Mann-Whitney U test and χ2 analysis. The performance of the deep learning image classification system resulted in accuracy of 78.2%, sensitivity of 75.4%, specificity of 81.0%, positive predictive value of 79.9%, negative predictive value of 77.1%, and area under the receiver operating characteristic curve of 0.80. These values were not significantly different from those found by the radiologists. The deep learning system yielded diagnostic results similar to those of the radiologists, which suggests that this system may be valuable for diagnostic support.

Alzheimer's disease diagnosis based on multiple cluster dense convolutional networks

Alzheimer's disease (AD) is an irreversible neurodegenerative disorder with progressive impairment of memory and cognitive functions. Structural magnetic resonance images (MRI) play important role to evaluate the brain anatomical changes for AD Diagnosis. Machine learning technologies have been widely studied on MRI computation and analysis for quantitative evaluation and computer-aided-diagnosis of AD. Most existing methods extract the hand-craft features after image processing such as registration and segmentation, and then train a classifier to distinguish AD subjects from other groups. Motivated by the success of deep learning in image classification, this paper proposes a classification method based on multiple cluster dense convolutional neural networks (DenseNets) to learn the various local features of MR brain images, which are combined for AD classification. First, we partition the whole brain image into different local regions and extract a number of 3D patches from each region. Second, the patches from each region are grouped into different clusters with the K-Means clustering method. Third, we construct a DenseNet to learn the patch features for each cluster and the features learned from the discriminative clusters of each region are ensembled for classification. Finally, the classification results from different local regions are combined to enhance final image classification. The proposed method can gradually learn the MRI features from the local patches to global image level for the classification task. There are no rigid registration and segmentation required for preprocessing MRI images. Our method is evaluated using T1-weighted MRIs of 831 subjects including 199 AD patients, 403 mild cognitive impairment (MCI) and 229 normal control (NC) subjects from Alzheimer's Disease Neuroimaging Initiative (ADNI) database. Experimental results show that the proposed method achieves an accuracy of 89.5% and an AUC (area under the ROC curve) of 92.4% for AD vs. NC classification, and an accuracy of 73.8% and an AUC of 77.5% for MCI vs. NC classification, demonstrating the promising classification performances.

Semantic enhanced deep learning for image classification

Semantic enhanced deep learning for image classification

Classification of Alzheimer's Disease by Combination of Convolutional and Recurrent Neural Networks Using FDG-PET Images.

Alzheimer’s disease (AD) is an irreversible brain degenerative disorder affecting people aged older than 65 years. Currently, there is no effective cure for AD, but its progression can be delayed with some treatments. Accurate and early diagnosis of AD is vital for the patient care and development of future treatment. Fluorodeoxyglucose positrons emission tomography (FDG-PET) is a functional molecular imaging modality, which proves to be powerful to help understand the anatomical and neural changes of brain related to AD. Most existing methods extract the handcrafted features from images, and then design a classifier to distinguish AD from other groups. These methods highly depends on the preprocessing of brain images, including image rigid registration and segmentation. Motivated by the success of deep learning in image classification, this paper proposes a new classification framework based on combination of 2D convolutional neural networks (CNN) and recurrent neural networks (RNNs), which learns the intra-slice and inter-slice features for classification after decomposition of the 3D PET image into a sequence of 2D slices. The 2D CNNs are built to capture the features of image slices while the gated recurrent unit (GRU) of RNN is cascaded to learn and integrate the inter-slice features for image classification. No rigid registration and segmentation are required for PET images. Our method is evaluated on the baseline FDG-PET images acquired from 339 subjects including 93 AD patients, 146 mild cognitive impairments (MCI) and 100 normal controls (NC) from Alzheimer’s Disease Neuroimaging Initiative (ADNI) database. Experimental results show that the proposed method achieves an area under receiver operating characteristic curve (AUC) of 95.3% for AD vs. NC classification and 83.9% for MCI vs. NC classification, demonstrating the promising classification performance.