Search results for: multi-temporal image classification

Authors: Hadeer R. M. Tawfik, Rania A. K. Birry, Amani A. Saad

Abstract:

Eyes are considered to be the most sensitive and important organ for human being. Thus, any eye disorder will affect the patient in all aspects of life. Cataract is one of those eye disorders that lead to blindness if not treated correctly and quickly. This paper demonstrates a model for automatic detection, classification, and grading of cataracts based on image processing techniques and artificial intelligence. The proposed system is developed to ease the cataract diagnosis process for both ophthalmologists and patients. The wavelet transform combined with 2D Log Gabor Wavelet transform was used as feature extraction techniques for a dataset of 120 eye images followed by a classification process that classified the image set into three classes; normal, early, and advanced stage. A comparison between the two used classifiers, the support vector machine SVM and the artificial neural network ANN were done for the same dataset of 120 eye images. It was concluded that SVM gave better results than ANN. SVM success rate result was 96.8% accuracy where ANN success rate result was 92.3% accuracy.

Keywords: cataract, classification, detection, feature extraction, grading, log-gabor, neural networks, support vector machines, wavelet

Procedia PDF Downloads 332

Authors: Akrem Sellami, Imed Riadh Farah

Abstract:

Hyperspectral imagery (HSI) typically provides a wealth of information captured in a wide range of the electromagnetic spectrum for each pixel in the image. Hence, a pixel in HSI is a high-dimensional vector of intensities with a large spectral range and a high spectral resolution. Therefore, the semantic interpretation is a challenging task of HSI analysis. We focused in this paper on object classification as HSI semantic interpretation. However, HSI classification still faces some issues, among which are the following: The spatial variability of spectral signatures, the high number of spectral bands, and the high cost of true sample labeling. Therefore, the high number of spectral bands and the low number of training samples pose the problem of the curse of dimensionality. In order to resolve this problem, we propose to introduce the process of dimensionality reduction trying to improve the classification of HSI. The presented approach is a semi-supervised band selection method based on spatial hypergraph embedding model to represent higher order relationships with different weights of the spatial neighbors corresponding to the centroid of pixel. This semi-supervised band selection has been developed to select useful bands for object classification. The presented approach is evaluated on AVIRIS and ROSIS HSIs and compared to other dimensionality reduction methods. The experimental results demonstrate the efficacy of our approach compared to many existing dimensionality reduction methods for HSI classification.

Keywords: dimensionality reduction, hyperspectral image, semantic interpretation, spatial hypergraph

Procedia PDF Downloads 306

Authors: Hung-Sheng Lin, Cheng-Hsuan Li

Abstract:

Over the past few years, kernel-based algorithms have been widely used to extend some linear feature extraction methods such as principal component analysis (PCA), linear discriminate analysis (LDA), and nonparametric weighted feature extraction (NWFE) to their nonlinear versions, kernel principal component analysis (KPCA), generalized discriminate analysis (GDA), and kernel nonparametric weighted feature extraction (KNWFE), respectively. These nonlinear feature extraction methods can detect nonlinear directions with the largest nonlinear variance or the largest class separability based on the given kernel function. Moreover, they have been applied to improve the target detection or the image classification of hyperspectral images. The double nearest proportion feature extraction (DNP) can effectively reduce the overlap effect and have good performance in hyperspectral image classification. The DNP structure is an extension of the k-nearest neighbor technique. For each sample, there are two corresponding nearest proportions of samples, the self-class nearest proportion and the other-class nearest proportion. The term “nearest proportion” used here consider both the local information and other more global information. With these settings, the effect of the overlap between the sample distributions can be reduced. Usually, the maximum likelihood estimator and the related unbiased estimator are not ideal estimators in high dimensional inference problems, particularly in small data-size situation. Hence, an improved estimator by shrinkage estimation (regularization) is proposed. Based on the DNP structure, LDA is included as a special case. In this paper, the kernel method is applied to extend DNP to kernel-based DNP (KDNP). In addition to the advantages of DNP, KDNP surpasses DNP in the experimental results. According to the experiments on the real hyperspectral image data sets, the classification performance of KDNP is better than that of PCA, LDA, NWFE, and their kernel versions, KPCA, GDA, and KNWFE.

Keywords: feature extraction, kernel method, double nearest proportion feature extraction, kernel double nearest feature extraction

Procedia PDF Downloads 344

Authors: Lae-Jeong Park, Kyusoo Chung, Jungho Moon

Abstract:

This paper introduces an effective method of segmenting Korean text (place names in Korean) from a Korean road sign image. A Korean advanced directional road sign is composed of several types of visual information such as arrows, place names in Korean and English, and route numbers. Automatic classification of the visual information and extraction of Korean place names from the road sign images make it possible to avoid a lot of manual inputs to a database system for management of road signs nationwide. We propose a series of problem-specific heuristics that correctly segments Korean place names, which is the most crucial information, from the other information by leaving out non-text information effectively. The experimental results with a dataset of 368 road sign images show 96% of the detection rate per Korean place name and 84% per road sign image.

Keywords: segmentation, road signs, characters, classification

Procedia PDF Downloads 444

Authors: Okike Benjamin, Garba E J. D.

Abstract:

Encryption is used to conceal information from prying eyes. Presently, information and data encryption are common due to the volume of data and information in transit across the globe on daily basis. Image encryption is yet to receive the attention of the researchers as deserved. In other words, video and multimedia documents are exposed to unauthorized accessors. The authors propose image encryption using matrix transpose. An algorithm that would allow image encryption is developed. In this proposed image encryption technique, the image to be encrypted is split into parts based on the image size. Each part is encrypted separately using matrix transpose. The actual encryption is on the picture elements (pixel) that make up the image. After encrypting each part of the image, the positions of the encrypted images are swapped before transmission of the image can take place. Swapping the positions of the images is carried out to make the encrypted image more robust for any cryptanalyst to decrypt.

Keywords: image encryption, matrices, pixel, matrix transpose

Procedia PDF Downloads 421

Authors: Jiahe Liu, Hongyang Yu, Miao Luo, Feng Qian

Abstract:

The widespread application of 3D human body reconstruction spans various fields. Smplify-X, an algorithm reliant on single-image input, employs three distinct body parameter templates, necessitating gender classification of individuals within the input image. Researchers employed a ResNet18 network to train a gender classifier within the Smplify-X framework, setting the threshold at 0.9, designating images falling below this threshold as having neutral gender. This model achieved 62.38% accurate predictions and 7.54% incorrect predictions. Our improvement involved refining the MobileNet network, resulting in a raised threshold of 0.97. Consequently, we attained 78.89% accurate predictions and a mere 0.2% incorrect predictions, markedly enhancing prediction precision and enabling more precise 3D human body reconstruction.

Keywords: SMPLX, mobileNet, gender classification, 3D human reconstruction

Procedia PDF Downloads 99

Authors: Kemal Polat

Abstract:

In this study, it has been estimated whether there is skin by using RBG values obtained from the camera and k-nearest neighbor (k-NN) classifier. The dataset used in this study has an unbalanced distribution and a linearly non-separable structure. This problem can also be called a big data problem. The Skin dataset was taken from UCI machine learning repository. As the classifier, we have used the k-NN method to handle this big data problem. For k value of k-NN classifier, we have used as 1. To train and test the k-NN classifier, 50-50% training-testing partition has been used. As the performance metrics, TP rate, FP Rate, Precision, recall, f-measure and AUC values have been used to evaluate the performance of k-NN classifier. These obtained results are as follows: 0.999, 0.001, 0.999, 0.999, 0.999, and 1,00. As can be seen from the obtained results, this proposed method could be used to predict whether the image is skin or not.

Keywords: k-NN classifier, skin or non-skin classification, RGB values, classification

Procedia PDF Downloads 248

Authors: R. Bouhennache, T. Bouden, A. A. Taleb, A. Chaddad

Abstract:

In this paper we propose a method to map the urban areas. The method uses an arithmetic calculation processed from the land features indexes and Tasseled cap transformation TC of multi spectral Thematic Mapper Landsat TM image. For this purpose the derived indexes image from the original image such SAVI the soil adjusted vegetation index, UI the urban Index, and EBBI the enhanced built up and bareness index were staked to form a new image and the bands were uncorrelated, also the Spectral Angle Mapper (SAM) and Spectral Information Divergence (SID) supervised classification approaches were first applied on the new image TM data using the reference spectra of the spectral library and subsequently the four urban, vegetation, water and soil land cover categories were extracted with their accuracy assessment.The urban features were represented using a logic calculation applied to the brightness, UI-SAVI, NDBI-greenness and EBBI- brightness data sets. The study applied to Blida and mentioned that the urban features can be mapped with an accuracy ranging from 92 % to 95%.

Keywords: EBBI, SAVI, Tasseled Cap Transformation, UI

Procedia PDF Downloads 482

Authors: N. Ghatwary, A. Ahmed, H. Jalab

Abstract:

In this paper, an approach for the liver tumor detection in computed tomography (CT) images is represented. The detection process is based on classifying the features of target liver cell to either tumor or non-tumor. Fractional differential (FD) is applied for enhancement of Liver CT images, with the aim of enhancing texture and edge features. Later on, a fusion method is applied to merge between the various enhanced images and produce a variety of feature improvement, which will increase the accuracy of classification. Each image is divided into NxN non-overlapping blocks, to extract the desired features. Support vector machines (SVM) classifier is trained later on a supplied dataset different from the tested one. Finally, the block cells are identified whether they are classified as tumor or not. Our approach is validated on a group of patients’ CT liver tumor datasets. The experiment results demonstrated the efficiency of detection in the proposed technique.

Keywords: fractional differential (FD), computed tomography (CT), fusion, aplha, texture features.

Procedia PDF Downloads 358

Authors: Yuntao Liu, Lei Wang, Haoran Xia

Abstract:

Machine learning has shown extensive applications in the development of classification models for autism spectrum disorder (ASD) using neural image data. This paper proposes a fusion multi-modal classification network based on a graph neural network. First, the brain is segmented into 116 regions of interest using a medical segmentation template (AAL, Anatomical Automatic Labeling). The image features of sMRI and the signal features of fMRI are extracted, which build the node and edge embedding representations of the brain map. Then, we construct a dynamically updated brain map neural network and propose a method based on a dynamic brain map adjacency matrix update mechanism and learnable graph to further improve the accuracy of autism diagnosis and recognition results. Based on the Autism Brain Imaging Data Exchange I dataset(ABIDE I), we reached a prediction accuracy of 74% between ASD and TD subjects. Besides, to study the biomarkers that can help doctors analyze diseases and interpretability, we used the features by extracting the top five maximum and minimum ROI weights. This work provides a meaningful way for brain disorder identification.

Keywords: autism spectrum disorder, brain map, supervised machine learning, graph network, multimodal data, model interpretability

Procedia PDF Downloads 66

Authors: Suzan Wedyan

Abstract:

Data mining is one of the main phases in the Knowledge Discovery Database (KDD) which is responsible of finding hidden and useful knowledge from databases. There are many different tasks for data mining including regression, pattern recognition, clustering, classification, and association rule. In recent years a promising data mining approach called associative classification (AC) has been proposed, AC integrates classification and association rule discovery to build classification models (classifiers). This paper surveys and critically compares several AC algorithms with reference of the different procedures are used in each algorithm, such as rule learning, rule sorting, rule pruning, classifier building, and class allocation for test cases.

Keywords: associative classification, classification, data mining, learning, rule ranking, rule pruning, prediction

Procedia PDF Downloads 537

Authors: Levent Basayigit, Sinan Demir, Yusuf Ucar, Burhan Kara

Abstract:

Many research projects require accurate delineation of the different land cover type of the agricultural area. Especially it is critically important for the definition of specific plants like cannabis. However, the complexity of vegetation stands structure, abundant vegetation species, and the smooth transition between different seconder section stages make vegetation classification difficult when using traditional approaches such as the maximum likelihood classifier. Most of the time, classification distinguishes only between trees/annual or grain. It has been difficult to accurately determine the cannabis mixed with other plants. In this paper, a mixed distribution models approach is applied to classify pure and mix cannabis parcels using Worldview-2 imagery in the Lakes region of Turkey. Five different land use types (i.e. sunflower, maize, bare soil, and cannabis) were identified in the image. A constrained Gaussian mixture discriminant analysis (GMDA) was used to unmix the image. In the study, 255 reflectance ratios derived from spectral signatures of seven bands (Blue-Green-Yellow-Red-Rededge-NIR1-NIR2) were randomly arranged as 80% for training and 20% for test data. Gaussian mixed distribution model approach is proved to be an effective and convenient way to combine very high spatial resolution imagery for distinguishing cannabis vegetation. Based on the overall accuracies of the classification, the Gaussian mixed distribution model was found to be very successful to achieve image classification tasks. This approach is sensitive to capture the illegal cannabis planting areas in the large plain. This approach can also be used for monitoring and determination with spectral reflections in illegal cannabis planting areas.

Keywords: Gaussian mixture discriminant analysis, spectral mixture model, Worldview-2, land parcels

Procedia PDF Downloads 197

Authors: Jonathan Sohn

Abstract:

Deep neural networks (DNNs) have shown state-of-the-art performance for many applications, including computer vision, natural language processing, and speech recognition. Recently, adversarial attacks have been studied in the context of deep neural networks, which aim to alter the results of deep neural networks by modifying the inputs slightly. For example, an adversarial attack on a DNN used for object detection can cause the DNN to miss certain objects. As a result, the reliability of DNNs is undermined by their lack of robustness against adversarial attacks, raising concerns about their use in safety-critical applications such as autonomous driving. In this paper, we focus on studying the adversarial attacks and defenses on DNNs for image classification. There are two types of adversarial attacks studied which are fast gradient sign method (FGSM) attack and projected gradient descent (PGD) attack. A DNN forms decision boundaries that separate the input images into different categories. The adversarial attack slightly alters the image to move over the decision boundary, causing the DNN to misclassify the image. FGSM attack obtains the gradient with respect to the image and updates the image once based on the gradients to cross the decision boundary. PGD attack, instead of taking one big step, repeatedly modifies the input image with multiple small steps. There is also another type of attack called the target attack. This adversarial attack is designed to make the machine classify an image to a class chosen by the attacker. We can defend against adversarial attacks by incorporating adversarial examples in training. Specifically, instead of training the neural network with clean examples, we can explicitly let the neural network learn from the adversarial examples. In our experiments, the digit recognition accuracy on the MNIST dataset drops from 97.81% to 39.50% and 34.01% when the DNN is attacked by FGSM and PGD attacks, respectively. If we utilize FGSM training as a defense method, the classification accuracy greatly improves from 39.50% to 92.31% for FGSM attacks and from 34.01% to 75.63% for PGD attacks. To further improve the classification accuracy under adversarial attacks, we can also use a stronger PGD training method. PGD training improves the accuracy by 2.7% under FGSM attacks and 18.4% under PGD attacks over FGSM training. It is worth mentioning that both FGSM and PGD training do not affect the accuracy of clean images. In summary, we find that PGD attacks can greatly degrade the performance of DNNs, and PGD training is a very effective way to defend against such attacks. PGD attacks and defence are overall significantly more effective than FGSM methods.

Keywords: deep neural network, adversarial attack, adversarial defense, adversarial machine learning

Procedia PDF Downloads 195

Authors: Fabio Fabris, Alex A. Freitas

Abstract:

Hierarchical classification is a special type of classification task where the class labels are organised into a hierarchy, with more generic class labels being ancestors of more specific ones. Meta-learning for classification-algorithm recommendation consists of recommending to the user a classification algorithm, from a pool of candidate algorithms, for a dataset, based on the past performance of the candidate algorithms in other datasets. Meta-learning is normally used in conventional, non-hierarchical classification. By contrast, this paper proposes a meta-learning approach for more challenging task of hierarchical classification, and evaluates it in a large number of bioinformatics datasets. Hierarchical classification is especially relevant for bioinformatics problems, as protein and gene functions tend to be organised into a hierarchy of class labels. This work proposes meta-learning approach for recommending the best hierarchical classification algorithm to a hierarchical classification dataset. This work’s contributions are: 1) proposing an algorithm for splitting hierarchical datasets into new datasets to increase the number of meta-instances, 2) proposing meta-features for hierarchical classification, and 3) interpreting decision-tree meta-models for hierarchical classification algorithm recommendation.

Keywords: algorithm recommendation, meta-learning, bioinformatics, hierarchical classification

Procedia PDF Downloads 314

Authors: Seynabou Toure, Oumar Diop, Kidiyo Kpalma, Amadou S. Maiga

Abstract:

Color and texture are the two most determinant elements for perception and recognition of the objects in an image. For this reason, color and texture analysis find a large field of application, for example in image classification and segmentation. But, the pioneering work in texture analysis was conducted on grayscale images, thus discarding color information. Many grey-level texture descriptors have been proposed and successfully used in numerous domains for image classification: face recognition, industrial inspections, food science medical imaging among others. Taking into account color in the definition of these descriptors makes it possible to better characterize images. Color texture is thus the subject of recent work, and the analysis of color texture images is increasingly attracting interest in the scientific community. In optical remote sensing systems, sensors measure separately different parts of the electromagnetic spectrum; the visible ones and even those that are invisible to the human eye. The amounts of light reflected by the earth in spectral bands are then transformed into grayscale images. The primary natural colors Red (R) Green (G) and Blue (B) are then used in mixtures of different spectral bands in order to produce RGB images. Thus, good color texture discrimination can be achieved using RGB under controlled illumination conditions. Some previous works investigate the effect of using different color space for color texture classification. However, the selection of the best performing color space in land-sea segmentation is an open question. Its resolution may bring considerable improvements in certain applications like coastline detection, where the detection result is strongly dependent on the performance of the land-sea segmentation. The aim of this paper is to present the results of a study conducted on different color spaces in order to show the best-performing color space for land-sea segmentation. In this sense, an experimental analysis is carried out using five different color spaces (RGB, XYZ, Lab, HSV, YCbCr). For each color space, the Haar wavelet decomposition is used to extract different color texture features. These color texture features are then used for Fusion of Over Segmentation (FOOS) based classification; this allows segmentation of the land part from the sea one. By analyzing the different results of this study, the HSV color space is found as the best classification performance while using color and texture features; which is perfectly coherent with the results presented in the literature.

Keywords: classification, coastline, color, sea-land segmentation

Procedia PDF Downloads 247

Authors: Kayne Uriel K. Rodrigo, Jerriane Hillary Heart S. Marcial, Samuel C. Brillo

Abstract:

Rice leaf diseases significantly impact yield production in rice-dependent countries, affecting their agricultural sectors. As part of precision agriculture, early and accurate detection of these diseases is crucial for effective mitigation practices and minimizing crop losses. Hence, this study proposes an enhancement to the Convolutional Neural Network (CNN), a widely-used method for Rice Leaf Disease Image Classification, by incorporating MobileViTV2—a recently advanced architecture that combines CNN and Vision Transformer models while maintaining fewer parameters, making it suitable for broader deployment on edge devices. Our methodology utilizes a publicly available rice disease image dataset from Kaggle, which was validated by a university structural biologist following the guidelines provided by the Philippine Rice Institute (PhilRice). Modifications to the dataset include renaming certain disease categories and augmenting the rice leaf image data through rotation, scaling, and flipping. The enhanced dataset was then used to train the MobileViTV2 model using the Timm library. The results of our approach are as follows: the model achieved notable performance, with 98% accuracy in both training and validation, 6% training and validation loss, and a Receiver Operating Characteristic (ROC) curve ranging from 95% to 100% for each label. Additionally, the F1 score was 97%. These metrics demonstrate a significant improvement compared to a conventional CNN-based approach, which, in a previous 2022 study, achieved only 78% accuracy after using 5 convolutional layers and 2 dense layers. Thus, it can be concluded that MobileViTV2, with its fewer parameters, outperforms traditional CNN models, particularly when applied to Rice Leaf Disease Image Identification. For future work, we recommend extending this model to include datasets validated by international rice experts and broadening the scope to accommodate biotic factors such as rice pest classification, as well as abiotic stressors such as climate, soil quality, and geographic information, which could improve the accuracy of disease prediction.

Keywords: convolutional neural network, MobileViTV2, rice leaf disease, precision agriculture, image classification, vision transformer

Procedia PDF Downloads 22

Authors: Felipe A. Guth, Shane Ward, Kevin McDonnell

Abstract:

The assessment of disease levels in crop fields is an important and time-consuming task that generally relies on expert knowledge of trained individuals. Image classification in agriculture problems historically has been based on classical machine learning strategies that make use of hand-engineered features in the top of a classification algorithm. This approach tends to not produce results with high accuracy and generalization to the classes classified by the system when the nature of the elements has a significant variability. The advent of deep convolutional neural networks has revolutionized the field of machine learning, especially in computer vision tasks. These networks have great resourcefulness of learning and have been applied successfully to image classification and object detection tasks in the last years. The objective of this work was to propose a new method based on deep learning convolutional neural networks towards the task of disease level monitoring. Common RGB images of winter wheat were obtained during a growing season. Five categories of disease levels presence were produced, in collaboration with agronomists, for the algorithm classification. Disease level tasks performed by experts provided ground truth data for the disease score of the same winter wheat plots were RGB images were acquired. The system had an overall accuracy of 84% on the discrimination of the disease level classes.

Keywords: crop disease assessment, deep learning, precision agriculture, residual neural networks

Procedia PDF Downloads 331

Authors: Jamshaid Ul Rahman, Faiza Makhdoom, Dianchen Lu

Abstract:

Activation functions play a decisive role in determining the capacity of Deep Neural Networks (DNNs) as they enable neural networks to capture inherent nonlinearities present in data fed to them. The prior research on activation functions primarily focused on the utility of monotonic or non-oscillatory functions, until Growing Cosine Unit (GCU) broke the taboo for a number of applications. In this paper, a Convolutional Neural Network (CNN) model named as ASU-CNN is proposed which utilizes recently designed activation function ASU across its layers. The effect of this non-monotonic and oscillatory function is inspected through feature map visualizations from different convolutional layers. The optimization of proposed network is offered by Adam with a fine-tuned adjustment of learning rate. The network achieved promising results on both training and testing data for the classification of CIFAR-10. The experimental results affirm the computational feasibility and efficacy of the proposed model for performing tasks related to the field of computer vision.

Keywords: amplifying sine unit, activation function, convolutional neural networks, oscillatory activation, image classification, CIFAR-10

Procedia PDF Downloads 111

Authors: Revoti Prasad Bora, Nikita Katyal, Saurabh Yadav

Abstract:

Pre-processing plays an important role in various image processing applications. Most of the time due to the similar nature of images, a particular pre-processing or a set of pre-processing steps are sufficient to produce the desired results. However, in the education domain, there is a wide variety of images in various aspects like images with line-based diagrams, chemical formulas, mathematical equations, etc. Hence a single pre-processing or a set of pre-processing steps may not yield good results. Therefore, a Deep Learning based approach for dynamically selecting a relevant pre-processing technique for each image is proposed. The proposed method works as a classifier to detect hidden patterns in the images and predicts the relevant pre-processing technique needed for the image. This approach experimented for an image similarity matching problem but it can be adapted to other use cases too. Experimental results showed significant improvement in average similarity ranking with the proposed method as opposed to static pre-processing techniques.

Keywords: deep-learning, classification, pre-processing, computer vision, image processing, educational data mining

Procedia PDF Downloads 163

Authors: Manoj Gupta, Nirmendra Singh Bhadauria

Abstract:

Most of the applications in image processing require high spatial and high spectral resolution in a single image. For example satellite image system, the traffic monitoring system, and long range sensor fusion system all use image processing. However, most of the available equipment is not capable of providing this type of data. The sensor in the surveillance system can only cover the view of a small area for a particular focus, yet the demanding application of this system requires a view with a high coverage of the field. Image fusion provides the possibility of combining different sources of information. In this paper, we have decomposed the image using DTCWT and then fused using average and hybrid of (maxima and average) pixel level techniques and then compared quality of both the images using PSNR.

Keywords: image fusion, DWT, DT-CWT, PSNR, average image fusion, hybrid image fusion

Procedia PDF Downloads 606

Authors: Zhong Ma, Zhuping Wang, Congxin Liu, Xiangzeng Liu

Abstract:

Satellite imagery classification is a challenging problem with many practical applications. In this paper, we designed a deep convolution neural network (DCNN) to classify the satellite imagery. The contributions of this paper are twofold — First, to cope with the large-scale variance in the satellite image, we introduced the inception module, which has multiple filters with different size at the same level, as the building block to build our DCNN model. Second, we proposed a genetic algorithm based method to efficiently search the best hyper-parameters of the DCNN in a large search space. The proposed method is evaluated on the benchmark database. The results of the proposed hyper-parameters search method show it will guide the search towards better regions of the parameter space. Based on the found hyper-parameters, we built our DCNN models, and evaluated its performance on satellite imagery classification, the results show the classification accuracy of proposed models outperform the state of the art method.

Keywords: satellite imagery classification, deep convolution network, genetic algorithm, hyper-parameter optimization

Procedia PDF Downloads 300

Authors: Jonathan Gong

Abstract:

Purpose AI-driven solutions are at the forefront of many pathology and medical imaging methods. Using algorithms designed to better the experience of medical professionals within their respective fields, the efficiency and accuracy of diagnosis can improve. In particular, X-rays are a fast and relatively inexpensive test that can diagnose diseases. In recent years, X-rays have not been widely used to detect and diagnose COVID-19. The under use of Xrays is mainly due to the low diagnostic accuracy and confounding with pneumonia, another respiratory disease. However, research in this field has expressed a possibility that artificial neural networks can successfully diagnose COVID-19 with high accuracy. Models and Data The dataset used is the COVID-19 Radiography Database. This dataset includes images and masks of chest X-rays under the labels of COVID-19, normal, and pneumonia. The classification model developed uses an autoencoder and a pre-trained convolutional neural network (DenseNet201) to provide transfer learning to the model. The model then uses a deep neural network to finalize the feature extraction and predict the diagnosis for the input image. This model was trained on 4035 images and validated on 807 separate images from the ones used for training. The images used to train the classification model include an important feature: the pictures are cropped beforehand to eliminate distractions when training the model. The image segmentation model uses an improved U-Net architecture. This model is used to extract the lung mask from the chest X-ray image. The model is trained on 8577 images and validated on a validation split of 20%. These models are calculated using the external dataset for validation. The models’ accuracy, precision, recall, f1-score, IOU, and loss are calculated. Results The classification model achieved an accuracy of 97.65% and a loss of 0.1234 when differentiating COVID19-infected, pneumonia-infected, and normal lung X-rays. The segmentation model achieved an accuracy of 97.31% and an IOU of 0.928. Conclusion The models proposed can detect COVID-19, pneumonia, and normal lungs with high accuracy and derive the lung mask from a chest X-ray with similarly high accuracy. The hope is for these models to elevate the experience of medical professionals and provide insight into the future of the methods used.

Keywords: artificial intelligence, convolutional neural networks, deep learning, image processing, machine learning

Procedia PDF Downloads 130

Authors: Farina Riaz, Shahab Abdulla, Srinjoy Ganguly, Hajime Suzuki, Ravinesh C. Deo, Susan Hopkins

Abstract:

Recently, quantum machine learning has received significant attention. For various types of data, including text and images, numerous quantum machine learning (QML) models have been created and are being tested. Images are exceedingly complex data components that demand more processing power. Despite being mature, classical machine learning still has difficulties with big data applications. Furthermore, quantum technology has revolutionized how machine learning is thought of, by employing quantum features to address optimization issues. Since quantum hardware is currently extremely noisy, it is not practicable to run machine learning algorithms on it without risking the production of inaccurate results. To discover the advantages of quantum versus classical approaches, this research has concentrated on colored image data. Deep learning classification models are currently being created on Quantum platforms, but they are still in a very early stage. Black and white benchmark image datasets like MNIST and Fashion MINIST have been used in recent research. MNIST and CIFAR-10 were compared for binary classification, but the comparison showed that MNIST performed more accurately than colored CIFAR-10. This research will evaluate the performance of the QML algorithm on the colored benchmark dataset CIFAR-10 to advance QML's real-time applicability. However, deep learning classification models have not been developed to compare colored images like Quantum Convolutional Neural Network (QCNN) to determine how much it is better to classical. Only a few models, such as quantum variational circuits, take colored images. The methodology adopted in this research is a hybrid approach by using penny lane as a simulator. To process the 10 classes of CIFAR-10, the image data has been translated into grey scale and the 28 × 28-pixel image containing 10,000 test and 50,000 training images were used. The objective of this work is to determine how much the quantum approach can outperform a classical approach for a comprehensive dataset of color images. After pre-processing 50,000 images from a classical computer, the QCNN model adopted a hybrid method and encoded the images into a quantum simulator for feature extraction using quantum gate rotations. The measurements were carried out on the classical computer after the rotations were applied. According to the results, we note that the QCNN approach is ~12% more effective than the traditional classical CNN approaches and it is possible that applying data augmentation may increase the accuracy. This study has demonstrated that quantum machine and deep learning models can be relatively superior to the classical machine learning approaches in terms of their processing speed and accuracy when used to perform classification on colored classes.

Keywords: CIFAR-10, quantum convolutional neural networks, quantum deep learning, quantum machine learning

Procedia PDF Downloads 129

Authors: Saad Alkentar, Abdulkareem Assalem

Abstract:

Recognizing small targets has always posed a significant challenge in image analysis. Over long distances, the image signal-to-noise ratio tends to be low, limiting the amount of useful information available to detection systems. Consequently, visual target recognition becomes an intricate task to tackle. In this study, we introduce a Track Before Detect (TBD) approach that leverages target trajectory information (coordinates) to effectively distinguish between noise and potential targets. By reframing the problem as a multivariate time series classification, we have achieved remarkable results. Specifically, our TBD method achieves an impressive 97% accuracy in separating target signals from noise within a mere half-second time span (consisting of 10 data points). Furthermore, when classifying the identified targets into our predefined categories—airplane, drone, and bird—we achieve an outstanding classification accuracy of 96% over a more extended period of 1.5 seconds (comprising 30 data points).

Keywords: small targets, drones, trajectory information, TBD, multivariate time series

Procedia PDF Downloads 47

Authors: Agastya Pratap Singh

Abstract:

Generative Adversarial Networks (GANs) have emerged as powerful tools in the fields of image recognition and anomaly detection due to their ability to model complex data distributions and generate realistic images. This systematic review explores recent advancements and applications of GANs in both image recognition and anomaly detection tasks. We discuss various GAN architectures, such as DCGAN, CycleGAN, and StyleGAN, which have been tailored to improve accuracy, robustness, and efficiency in visual data analysis. In image recognition, GANs have been used to enhance data augmentation, improve classification models, and generate high-quality synthetic images. In anomaly detection, GANs have proven effective in identifying rare and subtle abnormalities across various domains, including medical imaging, cybersecurity, and industrial inspection. The review also highlights the challenges and limitations associated with GAN-based methods, such as instability during training and mode collapse, and suggests future research directions to overcome these issues. Through this review, we aim to provide researchers with a comprehensive understanding of the capabilities and potential of GANs in transforming image recognition and anomaly detection practices.

Keywords: generative adversarial networks, image recognition, anomaly detection, DCGAN, CycleGAN, StyleGAN, data augmentation

Procedia PDF Downloads 20

Authors: Saleh Alshehri

Abstract:

Human skin detection is a vital step in many applications. Some of the applications are critical especially those related to security. This leverages the importance of a high-performance detection algorithm. To validate the accuracy of the algorithm, image databases are usually used. However, the suitability of these image databases is still questionable. It is suggested that the suitability can be measured mainly by the span the database covers of the color space. This research investigates the validity of three famous image databases.

Keywords: image databases, image processing, pattern recognition, neural networks

Procedia PDF Downloads 271

Authors: Ismail Elkhrachy

Abstract:

Determination of land use changing is an important component of regional planning for applications ranging from urban fringe change detection to monitoring change detection of land use. This data are very useful for natural resources management.On the other hand, the technologies and methods of change detection also have evolved dramatically during past 20 years. So it has been well recognized that the change detection had become the best methods for researching dynamic change of land use by multi-temporal remotely-sensed data. The objective of this paper is to assess, evaluate and monitor land use change surrounding the area of Najran city, Kingdom of Saudi Arabia (KSA) using Landsat images (June 23, 2009) and ETM+ image(June. 21, 2014). The post-classification change detection technique was applied. At last,two-time subset images of Najran city are compared on a pixel-by-pixel basis using the post-classification comparison method and the from-to change matrix is produced, the land use change information obtained.Three classes were obtained, urban, bare land and agricultural land from unsupervised classification method by using Erdas Imagine and ArcGIS software. Accuracy assessment of classification has been performed before calculating change detection for study area. The obtained accuracy is between 61% to 87% percent for all the classes. Change detection analysis shows that rapid growth in urban area has been increased by 73.2%, the agricultural area has been decreased by 10.5 % and barren area reduced by 7% between 2009 and 2014. The quantitative study indicated that the area of urban class has unchanged by 58.2 km〗^2, gained 70.3 〖km〗^2 and lost 16 〖km〗^2. For bare land class 586.4〖km〗^2 has unchanged, 53.2〖km〗^2 has gained and 101.5〖km〗^2 has lost. While agriculture area class, 20.2〖km〗^2 has unchanged, 31.2〖km〗^2 has gained and 37.2〖km〗^2 has lost.

Keywords: land use, remote sensing, change detection, satellite images, image classification

Procedia PDF Downloads 523

Authors: Nguyen J. M., Lucas G., Ruan S., Digonnet H., Antonioli D.

Abstract:

Introduction: Deep Learning (DL) is a very powerful tool for analyzing image data. But for tabular data, it cannot compete with machine learning methods like XGBoost. The research question becomes: can tabular data be transformed into images that can be analyzed by simple CNNs (Convolutional Neuron Networks)? Will DL be the absolute tool for data classification? All current solutions consist in repositioning the variables in a 2x2 matrix using their correlation proximity. In doing so, it obtains an image whose pixels are the variables. We implement a technology, DeepNIC, that offers the possibility of obtaining an image for each variable, which can be analyzed by simple CNNs. Material and method: The 'ROP' (Regression OPtimized) model is a binary and atypical decision tree whose nodes are managed by a new artificial neuron, the Neurop. By positioning an artificial neuron in each node of the decision trees, it is possible to make an adjustment on a theoretically infinite number of variables at each node. From this new decision tree whose nodes are artificial neurons, we created the concept of a 'Random Forest of Perfect Trees' (RFPT), which disobeys Breiman's concepts by assembling very large numbers of small trees with no classification errors. From the results of the RFPT, we developed a family of 10 statistical information criteria, Nguyen Information Criterion (NICs), which evaluates in 3 dimensions the predictive quality of a variable: Performance, Complexity and Multiplicity of solution. A NIC is a probability that can be transformed into a grey level. The value of a NIC depends essentially on 2 super parameters used in Neurops. By varying these 2 super parameters, we obtain a 2x2 matrix of probabilities for each NIC. We can combine these 10 NICs with the functions AND, OR, and XOR. The total number of combinations is greater than 100,000. In total, we obtain for each variable an image of at least 1166x1167 pixels. The intensity of the pixels is proportional to the probability of the associated NIC. The color depends on the associated NIC. This image actually contains considerable information about the ability of the variable to make the prediction of Y, depending on the presence or absence of other variables. A basic CNNs model was trained for supervised classification. Results: The first results are impressive. Using the GSE22513 public data (Omic data set of markers of Taxane Sensitivity in Breast Cancer), DEEPNic outperformed other statistical methods, including XGBoost. We still need to generalize the comparison on several databases. Conclusion: The ability to transform any tabular variable into an image offers the possibility of merging image and tabular information in the same format. This opens up great perspectives in the analysis of metadata.

Keywords: tabular data, CNNs, NICs, DeepNICs, random forest of perfect trees, classification

Procedia PDF Downloads 125

Authors: Ahmad Absetan, Mahdi Nooshyar

Abstract:

The importance map is an image-based measure and is a core part of the resizing algorithm. Importance measures include image gradients, saliency and entropy, as well as high level cues such as face detectors, motion detectors and more. In this work we proposed a new method to calculate the importance map, the importance map is generated automatically using a novel combination of image edge density and Harel saliency measurement. Experiments of different type images demonstrate that our method effectively detects prominent areas can be used in image resizing applications to aware important areas while preserving image quality.

Keywords: content-aware image resizing, visual saliency, edge density, image warping

Procedia PDF Downloads 582

Authors: Refilwe Moeletsi

Abstract:

Dimension stone quarrying around Brits and Belfast areas started in the early 1930s and has been growing rapidly since then. Environmental impacts associated with these quarries have not been documented, and hence this study aims at detecting any change in the environment that might have been caused by these activities. Landsat images that were used to assess land use/land cover changes in Brits quarries from 1998 - 2015. A supervised classification using maximum likelihood classifier was applied to classify each image into different land use/land cover types. Classification accuracy was assessed using Google Earth™ as a source of reference data. Post-classification change detection method was used to determine changes. The results revealed significant increase in granite quarries and corresponding decrease in vegetation cover within the study region.

Keywords: remote sensing, GIS, change detection, granite quarries

Procedia PDF Downloads 314