Figure 9 - uploaded by Deepan Perumal
Content may be subject to copyright.
Concepts of confusion matrix The working principle of confusion matrix processes is shown in figure 9. It is essential to find the confusion matrix while calculating the performance measures. Confusion matrix is a technique used to summarise results and used for validating classification methods. There are two common classes, which are usually deal with confusion matrix namely positive class and negative class. These two common classes can be further divided into four categories. True Positive is an outcome, where the model that has correct classification of positive example. False Negative is an outcome, where the model that has incorrect classification of positive examples. False Positive is an outcome, where the model that has incorrect classification of positive examples. True Negative is an outcome, where the model that has correct classification of negative examples.
Source publication
Remote sensing image (RSI) scene classification has
received growing attention from the research community in
recent days. Over the past few decades, with the rapid
development of deep learning models particularly
convolutional neural networks (CNN), the performance of
RSI scene classification have been drastically improved due
to the hierarchical...
Similar publications
Scheduling still constitutes a challenging problem, especially for complex problem settings involving due dates and sequence-dependent setups. The majority of existing approaches use heuristics or meta-heuristics, like Genetic Algorithms or Reinforcement Learning. We show that a supervised learning framework can learn and generalize from generated...
Super-resolution (SR) has become a widely researched topic in recent years. SR methods can improve overall image and video quality and create new possibilities for further content analysis. But the SR mainstream focuses primarily on increasing the naturalness of the resulting image despite potentially losing context accuracy. Such methods may produ...
Solidification of eutectic systems delivers compelling examples of microstructure formation, which makes the phenomena intriguing to many engineers and scientists. Therefore, eutectic solidification is extensively studied experimentally, theoretically, and numerically. However, some exotic eutectic microstructures are still not understood, particul...
The article presents the process of identifying discrete-continuous models with the use of heuristic algorithms. A stepped cantilever beam was used as an example of a discrete-continuous model. The theoretical model was developed based on the formalism of Lagrange multipliers and the Timoshenko theory. Based on experimental research, the theoretica...
Following the introduction of high-resolution player tracking technology, a new range of statistical analysis has emerged in sports, specifically in basketball. However, such high-dimensional data are often challenging for statistical inference and decision making. In this article we employ a state-of-the-art Bayesian mixture model that allows the...
Citations
... Figure 1 shows how to categorize an urban area image into commercial building, residential area, and industrial area. In general, different types of ground objects can be found in remote sensing images [2]. An industrial scene may contain roads, trees, buildings and so on. ...
... Research Article EAI Endorsed Transactions on Scalable Information Systems September 2022 -October 2022 | Volume 9 | Issue 5 | e5 P. Deepan et al. 2 complicated spatial ground objects that do not have a consistent shape and structure. The study of Simonyan and Zisserman [3] in Neural Information Processing Systems (NIPS) 2012 and success in the International Large Scale Visual Recognition Challenge (ILSVRC) 2012 ImageNet competition popularized the use of deep neural networks for image classification and recognition. ...
Remote Sensing Image (RSI) analysis has seen a massive increase in popularity over the last few decades, due to the advancement of deep learning models. A wide variety of deep learning models have emerged for the task of scene classification in remote sensing image analysis. The majority of these models have shown significant success. However, we found that there is significant variability, in order to improve the system efficiency in characterizing complex patterns in remote sensing imagery. We achieved this goal by expanding the architecture of VGG-16 Net and fine-tuning hyperparameters such as batch size, dropout probabilities, and activation functions to create the optimized Remote Sensing Image Scene Classification (RSISC-16 Net) deep learning model for scene classification. Using the Talos optimization tool, the results are carried out. This will increase efficiency and reduce the risk of over-fitting. Our proposed RSISC-16 Net model outperforms the VGG-16 Net model, according to experimental results.
... Taking the above disadvantages into consideration, our contribution is to propose a Mask Region based Dilated Convolutional Neural Network (Mask R-DCNN) for object detection of RSI images by using instance segmentation and object detection. The model efficiently increases the receptive fields of filer without loss of information in convolutional layer and also allows multi-class target detection in complex background situations [16]. The paper uses four backbone networks (ResNet-50, ResNet-101, Dilated ResNet-50 and Dilated ResNet-101) as feature extraction model. ...
Object detection in remote sensing image has received increasing attention from the research community in recent days. Over the past few decades, variety of deep learning based detection model such as Region based Convolutional Neural Network (R-CNN), Fast R-CNN and Faster R-CNN has been applied for object detection. However, most of the existing detection methods localize each object using the bounding box, but cannot segment the object from the background. So in order to tackle the issue, we introduce the Mask R-Dilated CNN model, which incorporates both object detection and segmentation. In Mask R-DCNN, ResNet-50 and ResNet-101 act as backbone for feature extraction, Region Proposal Network (RPN) is utilized to generate RoIs and RoIAlign is to carefully hold the exact spatial location to generate mask through Fully Convolution Network (FCN). The aim of Mask R-DCNN model is to incorporate more relevant information by increasing the receptive field of convolutional layer for improving the robustness. Experimental results on the NWPU VHR 10-class benchmark dataset demonstrated the effectiveness of the proposed model by providing 95.7% accuracy for Dilated ResNet-50 & 96.2% accuracy for Dilated ResNet-101, which is better than traditional Mask R-CNN model.
... Despite the recent breakthroughs of deep learning architectures in pattern recognition tasks, they need to estimate millions of parameters in the fully connected layers that require powerful hardware with high processing capacity and memory. Thus an efficient quantization based pooling method for face recognition using the VGG-16 pre-trained model is used [12,13]. At the last convolutional layer (also called channels) feature maps are represented using Bag-of-Features (BoF) paradigm. ...
Current scenario of COVID-19 (Corona Virus Disease) pandemic makes almost everyone to wear a mask in order to effectively prevent the spread of the virus. This almost makes conventional facial recognition technology ineffective in many cases, such as community access control, face access control, facial attendance, facial security checks at train stations, etc. Therefore, it is very urgent to improve the recognition performance of the existing face recognition technology on the masked faces. For that detecting the people with face mask is very essential. In this work, a reliable method based on discard masked region is proposed in order to address the problem of masked face recognition process. The first step is to discard the masked face region and extract the forehead and eyes region. Next, a pre-trained deep Convolutional neural networks (CNN) is applied to extract the best features from the obtained regions. Finally, it show experimental result is achieved an accuracy of more than 98% validation dataset.
... With the high level development of profound learning, Convolutional Neural Network (CNN) model turns into a novel methodology for addressing highlights and fundamentally improved the presentation of article location [12,13]. The CNN based article discovery calculations can be ordered into two classifications in particular locale based CNN strategies which incorporates R-CNN, Fast R-CNN, Faster R-CNN and district free techniques which incorporates SSD and YOLO. ...
Aircraft detection in remote sensing images (RSI) is receiving significant attention from researchers in recent years as it is a challenging problem in satellite image analysis field and has wide range of applications. From literature survey, we found that complex backgrounds, vertical views and variation in size are the factors that affect the speed and accuracy of an algorithm. In this paper, we have used You Only Look Only (YOLOv3) model which is a state-of-the art object detection algorithm and is fast and accurate than Region-based Convolutional Neural Network (R-CNN), Fast R-CNN, Faster R-CNN and Single Shot Multi-box Detector (SSD). With the help of multi-scale detection and Non-max suppression techniques (NMS) available in YOLOv3, we have successfully eliminated overlapping bounding boxes and detected aircrafts of varying size which is a challenging task in object detection algorithms. We have evaluated our approach on two public datasets namely NWPU 45-class and PatternNet dataset and demonstrated that it achieves relatively good detection performance.
Convolutional Neural Networks (CNNs) have emerged as a powerful tool for image classification tasks, owing to their ability to automatically learn hierarchical representations from raw pixel data. This paper presents a comprehensive analysis of CNNs for image classification, focusing on their architecture, training process, and performance evaluation metrics. The study investigates various CNN architectures, including AlexNet, GoogLeNet, and ResNet, highlighting their respective strengths and weaknesses in different scenarios. Moreover, the paper explores the impact of hyperparameters such as network depth, filter size, and pooling strategies on classification accuracy and computational efficiency. Furthermore, the training process of CNNs, encompassing data preprocessing, augmentation techniques, and optimization algorithms, is scrutinized to elucidate best practices for achieving optimal performance. Additionally, the paper discusses commonly used evaluation metrics such as accuracy, precision, recall, and F1-score, elucidating their interpretation and significance in assessing model performance. Through a systematic review of existing literature and experimental validation, this analysis aims to provide insights into the underlying mechanisms of CNNs for image classification tasks. Finally, the paper outlines future research directions, emphasizing the need for exploring novel architectures, optimizing hyperparameters, and enhancing interpretability and robustness of CNN models. Overall, this study contributes to a deeper understanding of CNNs for image classification and provides valuable guidance for practitioners and researchers in the field.
Classification of remote sensing images (RSIs) is a challenging task and has become an active research topic in the field of remote sensing community. Over the past six decades, variety of machine learning algorithms such as logistic regression (LR), K-nearest neighbours (K-NN), random forest (RF), support vector machine (SVM) and multilayer perceptron (MLP) has been applied for scene classification. In order to improve robustness over a single model, we have introduced a hybrid approach called as ensembling which is nothing but training multiple models instead of a single model and to combine predictions from these models. Five different ensemble methods, namely AdaBoost, bagging, majority voting, weighted voting and stacking, are evaluated in this paper. For evaluating the proposed approach, we have collected 8000 remote sensing images from PatternNet dataset and found that ensembling majority voting technique applied with MLP, SVM-linear, SVM-kernel and RF classifiers shows an out performance of 93.5% accuracy which is higher than the individual classifiers.
