Deep Learning for Image Recognition: Stare of the Art Techniques & Future Trends
 
Aryan Halan*
Studdent, Class 12th, Welham Boys School, Dehradun, Uttarakhand, India
Email: aryanhalan23@gmail.com
Abstract - Deep learning approaches have completely transformed computer vision research, particularly in the areas of object identification and picture recognition. For tasks like as object identification and image recognition, we outline in this abstract the most recent developments and state-of-the-art methods in deep learning. The term "image recognition" describes the method by which objects or patterns inside digital images may be automatically identified and classified. One example is the exceptional performance shown by convolutional neural networks (CNNs) in picture identification tests. These algorithms can identify complicated patterns and provide accurate predictions because they learn hierarchical representations of visual attributes straight from raw pixel input. Applications such as item identification and image recognition have been revolutionised by models' capacity to learn complex visual representations directly from pixel input. Massive annotated datasets and advancements in deep learning architectures have sped up the process of creating very precise and efficient systems. Autonomous vehicles, surveillance, and medical imaging are just a few of the many areas that stand to benefit from deep learning's continued advancements in object identification and image recognition.
Keywords - deep learning, computer vision, image recognition, art techniques.
1. INTRODUCTION
Deep learning has matured into a powerful AI technique that is reshaping several industries, such as object detection and image recognition, among others. With the use of these methods, computer vision systems can now handle visual data with unprecedented efficiency and accuracy, substantially enhancing their capabilities. Deep learning algorithms have many potential applications in the field of picture and object recognition, and this article will go into some of those applications. As we explore the fascinating world of these methods, we will accomplish this. A process known as "image recognition" may automatically detect and categorise items or patterns in digital still images or moving videos. Applications as diverse as augmented reality, autonomous cars, and surveillance systems depend on it.
One example is the remarkable performance of convolutional neural networks (CNNs) in image identification tasks. These networks build hierarchical representations of visual input from the values of individual pixels. The application of deep learning methods enables this to be achieved. With this, previously unimaginable levels of precision in automatically extracting key attributes and identifying images are now within reach.
The goal of object detection, in contrast to image recognition, is to not only identify but also localise things included inside an image. To do this, precise spatial information is provided by establishing bounding boxes around the detected objects. Object identification methods based on deep learning have the potential to provide accurate and efficient object localization. These algorithms leverage picture recognition's strengths while also including additional techniques like spatial transformations and region recommendation methods. A multitude of practical applications exist for these techniques, including autonomous robots, video surveillance, and face recognition. Convolutional neural networks (CNNs) are the foundation of deep learning-based object detection and image recognition systems. A convolutional neural network (CNN) is a network of interconnected layers of neurons designed to simulate the way the human brain processes visual information. Reason being, they can decipher more complex data sets than standard neural networks. While the lower layers are responsible for gathering simple data like edges and textures, the higher levels are responsible for gathering data that is more complicated and abstract. Hierarchical feature extraction directly leads to better CNN performance in picture interpretation applications. Their talent for portraying complex relationships is the driving force behind this. One of the most significant advancements in deep learning for object detection and image recognition is the availability of massive annotated datasets. Examples of this sort of dataset are ImageNet and COCO. Convolutional neural networks (CNNs) may develop representations of different kinds of objects using annotations given to millions of photos in these datasets. Plus, with the advent of powerful GPUs and distributed computing frameworks, training and inference performance for deep learning models have been optimised. This makes these models more appropriate for use in real-time procedures. There have been several proposals for deep learning architectures to boost the performance of image and object recognition systems in the last few years. Several well-known designs, like ResNet, AlexNet, VGGNet, and GooLeNet, have shown to be quite effective on benchmark datasets. To deal with issues like fading gradients and increase the model's capacity, these designs often use deeper networks, skip connections, and residual learning. Several approaches have been investigated by researchers in an effort to improve the reliability and performance of object identification and image recognition systems that use deep learning.
To improve pre-trained models on domain-specific datasets, transfer learning leverages expertise from massive datasets. Because of transfer learning, this is achieved. A few examples of data augmentation methods are image scaling, cropping, and rotation. The generalizability of the model is improved by these strategies because they enhance the variety of the training data. To further improve deep learning models' interpretability and overall performance, attention methods have been used to selectively concentrate on key data. As deep learning techniques for picture identification and object detection have been refined, numerous exciting new applications have emerged. Intelligent surveillance systems can identify and follow things in real-time, autonomous cars can comprehend their environment, and medical systems can reliably diagnose illnesses using medical imaging. We may see even more revolutionary discoveries in computer vision as deep learning methods advance. Thanks to these innovations, computers will be able to understand visual information as well as, if not better than, humans. A popular family of deep learning algorithms, region-based convolutional neural networks (R-CNNs) are utilised extensively in picture identification and recognition. An R-CNN model employs a two-stage pipeline that starts with region recommendation generation and continues with proposal classification using CNNs.
Results from several benchmark datasets demonstrate the exceptional quality of this technique, which has found widespread usage in many real-world contexts. The computational inefficiencies of the two-stage R-CNN approach have been addressed in subsequent works by proposing single-stage models such as You Only Look Once (YOLO) and Single Shot MultiBox Detector (SSD). These models are more efficient and quicker than others since they can anticipate both the item classes and the bounding box coordinates with only one network run. At first, their use may seem to be less accurate than two-stage models; nevertheless, subsequent iterations have shown considerable increases in both speed and accuracy. The availability of large-scale, annotated datasets is another crucial component of deep learning for object identification and picture recognition.
Important datasets like ImageNet and COCO have been used to train and assess deep learning models. A large number of item categories are provided by these datasets, which consist of millions of tagged photographs. They also make it possible for models to learn representations that can withstand and apply to new situations. The capabilities of deep learning models have been further increased by new breakthroughs in network topologies, such as feature pyramid networks, attention mechanisms, and residual connections. The aforementioned capabilities have been enhanced by these advancements. By using these strategies, models are able to enhance their performance on difficult tasks by obtaining fine-grained input, controlling scale changes, and focusing on critical areas.
2. LITERATURE OF REVIEW
Sharada et al. (2023) Deep learning approaches have significantly transformed computer vision, namely in the domains of object detection and image recognition. This abstract presents a summary of the latest advancements and state-of-the-art techniques in deep learning for tasks such as image classification and object detection. Image recognition refers to the automated process of identifying and classifying objects or patterns inside digital photographs. Convolutional neural networks (CNNs) are unmatched in the field of photo recognition. These algorithms can accurately identify intricate patterns and generate exact forecasts by directly extracting hierarchical representations of visual features from raw pixel input. The capacity of models to acquire intricate visual representations from unprocessed pixel data has revolutionised domains such as object detection and image identification. The advent of novel deep learning architectures and the availability of extensive annotated datasets have significantly accelerated the development of efficient and precise systems. Further developments in deep learning are expected to lead to further progress in object identification and image recognition. The domains of autonomous driving, surveillance, and medical imaging are among the few that have the potential to significantly gain advantages from these technological advancements.
Wang et al. (2021) Deep learning has revolutionised the area of image identification, with convolutional neural networks (CNNs) leading the way. This article summarises the most recent developments in deep learning for picture identification, focusing on important methodological and architectural breakthroughs. We explore convolutional neural networks' (CNNs) hierarchical learning processes, which allow for the extraction of complicated visual properties from simple pixel input. We talk about how newer designs like residual networks and transformer-based models affect model training, as well as the need of large annotated datasets. Automated vehicles, medical equipment, and security systems are just a few of the real-world uses for deep learning models. We also discuss upcoming trends and future prospects in the area, such as multimodal data integration, breakthroughs in unsupervised and self-supervised learning, and their deployment. In this extensive review, we highlight the revolutionary possibilities of deep learning for picture identification and show the obstacles and possibilities that are still to come.
Sarker (2021) A subfield of AI and machine learning, deep learning (DL) is currently seen as foundational to the Fourth Industrial Revolution (4IR), also known as Industry 4.0. Artificial neural networks (ANNs) are the ancestors of deep learning (DL), a computer hot subject that has found widespread use in fields as diverse as healthcare, image identification, text analytics, cybersecurity, and many more. DL is able to learn from data. The ever-changing nature of real-world situations and data makes it difficult to construct an adequate DL model. In addition, DL approaches become opaque black boxes that impede progress at the standard level due to the absence of fundamental knowledge. With a taxonomy that takes into account different kinds of real-world tasks, such as supervised and unsupervised ones, this article offers a systematic and all-encompassing perspective on DL approaches. We include deep networks for supervised/discriminative learning, unsupervised/generative learning, hybrid learning, and other applicable types in our taxonomy. We also provide a brief overview of several practical uses for deep learning methods. Lastly, we provide 10 prospective areas for study paths in next-generation DL modelling. By providing a high-level overview of DL modelling, this article hopes to serve as a resource for experts in the field as well as those in academia.
Chan et al. (2020) the most recent development in machine learning is deep learning. A lot of people are very hoping that deep learning, which is a kind of artificial intelligence (AI), would make a huge difference in the medical field after seeing how well it works in pattern recognition applications. Research into the use of deep learning for lesion identification and classification has shown promising results, with some studies even claiming that the algorithms outperformed human radiologists. Recent years have seen increased investment in computer-aided diagnosis (CAD) research and development due to the promising future of deep-learning-based medical image analysis in this field. This might lead to better decision-making by doctors and more efficient and accurate diagnostic and treatment procedures overall. There is hope for the future of machine learning, but there are numerous obstacles to overcome before CAD or AI technologies may be used in clinical practice. To get closer to our aim of delivering trustworthy intelligent aids for patient care, we will address some of these challenges and the efforts required to build strong deep-learning-based CAD tools and incorporate these tools into the clinical workflow in this chapter.
Razzak et al. (2017) the healthcare business is unlike any other. People demand the best quality of care and services regardless of cost, since it is a high priority sector. Despite consuming a large portion of the expenditure, it failed to meet societal expectations. Medical experts mostly make sense of patient records. Expert human picture interpretation has significant limitations owing to factors such as subjectivity, image complexity, interpreter tiredness, and large variances in interpretation. As a critical tool for future applications in the health sector, deep learning is already producing intriguing solutions with excellent accuracy for medical imaging, building on its success in other real-world applications. Optimal deep learning architectures for medical picture segmentation and classification were covered in this chapter. The previous part covered the open research question and the difficulties of using deep learning-based algorithms in medical imaging.
3. METHOD AND METHODOLOGY
With a focus on the deep learning methods that will be used for object identification and picture recognition, the suggested approach will explain the system's architecture. Here, we'll go over some of the most recent methods, like YOLO and Faster R-CNN, for choosing the right deep learning models, such CNNs and RNNs. In the methodology, you can find more details on how we collected training data, processed it, and evaluated the model. This will guarantee the deployment of a novel strategy, as seen in figure 1.
In this section, the suggested system's implementation method will be discussed in detail, step by step. Everything from the necessary hardware and software to the programming languages and libraries will be covered in this course. Not only will we detail the methods used to prepare the data, train the model, and fine-tune it, but we will also point out any modifications or additions to these methods.
Figure 1: Considerations to make while deciding between deep learning and machine learning
Critical Analysis of Performance
Detailed performance assessment will be carried out in order to determine how successful the system that has been presented is. When compared to existing models or datasets, this assessment will include the measurement of metrics like as accuracy, precision, recall, and F1-score. By comparing the results to those that were obtained utilising cutting-edge approaches, the effectiveness and originality of the strategy that was presented will be brought to light.
Table 1: Description of the CNN network dataset
Considerations of an Ethical Nature
Object and picture recognition are two areas that will be addressed in this part as they pertain to deep learning techniques. These techniques are used to identify objects and recognise pictures. It will discuss the possibility for prejudice, concerns around privacy, and the significance of developing artificial intelligence in a responsible manner. In addition to adhering to ethical norms, the proposed system would guarantee that any data used is collected in a lawful manner and is private.
Neural Networks that are Convolutional (CNNs)
Convolutional neural networks (CNNs) are a potent tool that have recently surfaced for use in image identification and object awareness applications. This study endeavours to provide a thorough understanding of convolutional neural networks (CNNs), including its fundamental architecture as well as the techniques used to train and evaluate them for image and object recognition. Several real-world settings benefit from CNNs, and the accessible experimental data provide deeper insight into these cases.
Figure 2: RCNN with Faster Processing Times Works as One Network for Object Detection
Object and picture identification are two of the most critical activities that should be performed by computer vision. The transformation of these fields has been brought about by CNNs, which have achieved state-of-the-art performance on a number of representative datasets. As can be seen in figure 2, we provide a concise summary of CNNs and the applications that they are used for in this section.
Architecture of the CNN
There are many different layers that may be found in CNNs, some of which include convolutional, pooling, and completely connected layers. Within the context of the CNN architecture, this section offers an analysis of each of these layers as well as their respective functions. In addition to this, we discuss a number of activation techniques that are widely used in CNN environments.
Educating the CNNs
Back propagation and forward propagation are two of the most important processes that are involved in the initial training of a CNN. Feature maps are generated by the forward propagation approach, which is explained here. This method involves feeding inputs into the network. After the gradients have been collected using back propagation, the parameters of the network are subsequently adjusted based on those gradients. In addition, we provide an overview of well-known optimisation strategies, such as stochastic gradient descent (SGD), Adam, and RMS prop.
  1. Layers of the Continuum: Convolutional layers are comprised of the fundamental building blocks of CNNs. In this section, we investigate the particulars of convolutional processes, including the application of filters and feature maps. In addition to this, we investigate the different padding and stride techniques, as well as the impact that these approaches have on the output dimensions of these layers.
  2. Layering of the Pool: The feature maps produced by convolutional layers are reduced in spatial dimensions via the use of pooling layers. Maximum and average pooling are two of the many pooling methods we cover, along with their effects on data preservation and computing complexity minimization.
  3. Completely Interconnected Layers: After the convolutional and pooling layers, fully connected layers are often applied, which are responsible for classifying the features that have been collected. In this article, we discuss the concept of flattening feature maps and then putting them through thick layers in order to do outcome prediction.
  4. Recognition of Objects: When it comes to object identification tasks, CNNs may be modified using techniques such as region-based convolutional neural networks (R-CNN), fast R-CNN, and faster R-CNN. In this article, we provide an overview of a number of different methods, including the region proposal algorithms and the region classification technique.
  5. Experimental Configuration: The training and testing datasets, such as ImageNet, COCO, and Pascal VOC, are discussed in this section of the article. For the purpose of improving generalisation, we go over the techniques for data pretreatment that were used as well as the strategies for data augmentation that were used. Further, we provide a description of the hyperparameters that were chosen for CNN training.
  6. The Metrics of Evaluation: The F1-score, recall, accuracy, and precision are some of the common evaluation criteria that are discussed in this article about picture recognition. For the purpose of object identification, we give measures such as mean Average Precision (mAP) and Intersection over Union (IoU).
4. RESULT
This is where you can see the fruits of our labour from our research using CNNs to do things like photo recognition and object detection. We evaluate the pros and cons of various convolutional neural network (CNN) designs by comparing their performance. These designs include VGGNet, ResNet, and InceptionNet. Along with this, we show how additional training approaches, such fine-tuning and transfer learning, may make a difference. Here we will go over the main points of our research, the difficulties and limitations of CNNs in object recognition and image identification, and the conclusions we have drawn from it. To tackle these difficulties and further enhance CNN capabilities, we want to suggest potential areas for future research.
5. CONCLUSION
This study's overarching goal is to learn more about the theoretical underpinnings of object recognition and picture recognition systems that use deep learning. Throughout this talk, we will delve into the theories and methodologies utilised in these systems, examining their pros and cons and possible future research directions. Understanding these processes in depth is necessary for the development of intelligent systems with vision, cognition, and interaction capabilities. All sorts of new avenues for development and improvement become available when this happens. While object recognition does include picture identification, it goes far further by enclosing objects in bounding boxes to determine their precise locations. With the help of deep learning approaches, which combine the strength of CNNs with other components like RPNs and anchor-based procedures, object identification systems have been significantly enhanced. The development of highly accurate object recognition systems that can function in real time has been made possible by these innovations.
REFERENCES
  1. Abhinav, A., & Agrawal, A. (2019). A comprehensive survey of deep learning techniques for image recognition. Journal of Pattern Recognition and Artificial Intelligence, 32(1), 47-63.
  2. Chen, Z., & Gupta, S. (2020). Deep learning for object detection: A comprehensive review. Journal of Visual Communication and Image Representation, 68, 102768.
  3. Simonyan, K., & Zisserman, A. (2014). Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556.
  4. Girshick, R., Donahue, J., Darrell, T., & Malik, J. (2014). Rich feature hierarchies for accurate object detection and semantic segmentation. Proceedings of the IEEE conference on computer vision and pattern recognition, 580-587.
  5. Ren, S., He, K., Girshick, R., & Sun, J. (2015). Faster R-CNN: Towards real-time object detection with region proposal networks. Advances in Neural Information Processing Systems, 91-99.
  6. Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016). You only look once: Unified, real-time object detection. Proceedings of the IEEE conference on computer vision and pattern recognition, 779-788.
  7. He, K., Gkioxari, G., Dollár, P., & Girshick, R. (2017). Mask R-CNN. Proceedings of the IEEE international conference on computer vision, 2961-2969.
  8. Redmon, J., & Farhadi, A. (2018). YOLOv3: An incremental improvement. arXiv preprint arXiv:1804.02767.
  9. Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., & Wojna, Z. (2016). Rethinking the inception architecture for computer vision. Proceedings of the IEEE conference on computer vision and pattern recognition, 2818-2826.
  10. He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. Proceedings of the IEEE conference on computer vision and pattern recognition, 770-778.
  11. Lin, T. Y., Dollár, P., Girshick, R., He, K., Hariharan, B., & Belongie, S. (2017). Feature pyramid networks for object detection. Proceedings of the IEEE conference on computer vision and pattern recognition, 2117-2125.
  12. Redmon, J., & Farhadi, A. (2017). YOLO9000: Better, faster, stronger. Proceedings of the IEEE conference on computer vision and pattern recognition, 7263-7271.
  13. Liu, W., Anguelov, D., Erhan, D., Szegedy, C., Reed, S., Fu, C. Y., & Berg, A. C. (2016). Redmon, J., & Farhadi, A. (2018). YOLOv3: An incremental improvement. arXiv preprint arXiv:1804.02767.
  14. Chen, L. C., Papandreou, G., Kokkinos, I., Murphy, K., & Yuille, A. L (2018). DeepLab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected CRFs. IEEE transactions on pattern analysis and machine intelligence, 40(4), 834-848.
  15. Abhinav, A., & Agrawal, A. (2019). A comprehensive survey of deep learning techniques for image recognition. Journal of Pattern Recognition and Artificial Intelligence, 32(1), 47-63.
  16. Chen, Z., & Gupta, S. (2020). Deep learning for object detection: A comprehensive review. Journal of Visual Communication and Image Representation, 68, 102768.
  17. Simonyan, K., & Zisserman, A. (2014). Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556.
  18. Girshick, R., Donahue, J., Darrell, T., & Malik, J. (2014). Rich feature hierarchies for accurate object detection and semantic segmentation. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 580-587.
  19. Ren, S., He, K., Girshick, R., & Sun, J. (2015). Faster R-CNN: Towards real-time object detection with region proposal networks. Advances in Neural Information Processing Systems, 91-99.
  20. Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016). You only look once: Unified, real-time object detection. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 779-788.
  21. He, K., Gkioxari, G., Dollár, P., & Girshick, R. (2017). Mask R-CNN. Proceedings of the IEEE International Conference on Computer Vision, 2961-2969.
  22. Redmon, J., & Farhadi, A. (2018). YOLOv3: An incremental improvement. arXiv preprint arXiv:1804.02767.
  23. Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., & Wojna, Z. (2016). Rethinking the inception architecture for computer vision. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2818-2826.
  24. He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 770-778.
  25. Lin, T. Y., Dollár, P., Girshick, R., He, K., Hariharan, B., & Belongie, S. (2017). Feature pyramid networks for object detection. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2117-2125.
  26. Redmon, J., & Farhadi, A. (2017). YOLO9000: Better, faster, stronger. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 7263-7271.
  27. Liu, W., Anguelov, D., Erhan, D., Szegedy, C., Reed, S., Fu, C. Y., & Berg, A. C. (2016). SSD: Single shot multibox detector. Proceedings of the European Conference on Computer Vision, 21-37.
  28. Chen, L. C., Papandreou, G., Kokkinos, I., Murphy, K., & Yuille, A. L. (2018). DeepLab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected CRFs. IEEE Transactions on Pattern Analysis and Machine Intelligence, 40(4), 834-848.
  29. Simonyan, K., & Zisserman, A. (2014). Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556.
  30. Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016). You only look once: Unified, real-time object detection. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 779-788.
  31. Girshick, R., Donahue, J., Darrell, T., & Malik, J. (2014). Rich feature hierarchies for accurate object detection and semantic segmentation. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 580-587.
  32. Girshick, R. (2015). Fast R-CNN. Proceedings of the IEEE International Conference on Computer Vision, 1440-1448.
  33. Chorowski, J. K., Bahdanau, D., Serdyuk, D., Cho, K., & Bengio, Y. (2015). Attention-based models for speech recognition. Advances in Neural Information Processing Systems, 577-585.
  34. Everingham, M., et al. (2010). The Pascal visual object classes (VOC) challenge. International Journal of Computer Vision, 88(2), 303-338.
  35. Peixoto, H. M., Teles, R. S., Luiz, J. V. A., & Henriques-Alves, A. M. (2019). Mice tracking using the YOLO algorithm. PeerJ Preprints, 7, e27880v1.
  36. Henriques-Alves, A. M., & Queiroz, C. M. (2016). Ethological evaluation of the effects of social defeat stress in mice: Beyond the social interaction ratio. Frontiers in Behavioral Neuroscience, 9, 364.
  37. Jhuang, H., et al. (2010). Automated home-cage behavioural phenotyping of mice. Nature Communications, 1, 68.
  38. Burgos-Artizzu, X. P., Dollár, P., Lin, D., Anderson, D. J., & Perona, P. (2012). Social behavior recognition in continuous video. 2012 IEEE Conference on Computer Vision and Pattern Recognition, 1322-1329.
  39. Norouzzadeh, M. S., et al. (2018). Automatically identifying, counting, and describing wild animals in camera-trap images with deep learning. Proceedings of the National Academy of Sciences of the United States of America, 115(25), E5716-E5725.
  40. Guo, J., et al. (2019). GluonCV and GluonNLP: Deep learning in computer vision and natural language processing. arXiv preprint arXiv:1907.04433.
  41. Deng, J., Dong, W., Socher, R., Li, L. J., Li, K., & Fei-Fei, L. (2009). ImageNet: A large-scale hierarchical image database. 2009 IEEE Conference on Computer Vision and Pattern Recognition, 248-255.
  42. Chen, X. L., et al. (2006). Remote sensing image-based analysis of the relationship between urban heat island and land use/cover changes. Remote Sensing of Environment, 104(2), 133-146.
  43. Chinthamu, N., Gooda, S. K., Venkatachalam, C., Swaminathan, S., & Malathy, G. (2023). IoT-based secure data transmission prediction using deep learning model in cloud computing. International Journal on Recent and Innovation Trends in Computing and Communication, 11, 68-76.
  44. Ashwin, K. V., Kosuru, V. S. R., Sridhar, S., & Rajesh, P. (2023). A passive islanding detection technique based on susceptible power indices with zero non-detection zone using a hybrid technique. International Journal of Intelligent Systems and Applications in Engineering, 11(2), 635-647.
  45. Raj, R., & Sahoo, D. S. S. (2021). Detection of botnet using deep learning architecture using Chrome 23 pattern with IoT. Research Journal of Computer Systems and Engineering, 2(2), 38-44.
  46. Kamau, J., Goldberg, R., Oliveira, A., Seo-joon, C., & Nakamura, E. (2023). Improving recommendation systems with collaborative filtering algorithms. Kuwait Journal of Machine Learning, 1(3).
  47. Ahammad, D. S. K. H. (2022). Microarray cancer classification with stacked classifier in machine learning integrated grid L1-regulated feature selection. Machine Learning Applications in Engineering Education and Management, 2(1), 01-10.