Home and industrial intelligent automation systems are now possible as a result of the rapid advancement in artificial intelligence (AI) and robotics. A design and implementation of an AI-controlled robot arm that automatically recognizes, classifies, and sorts objects by their colors are discussed in this paper. The system uses a camera and OpenCV for grabbing live video streams and HSV thresholding to recognize the colors of the objects. An Arduino microcontroller receives detected color information, decodes command, and employs an L293D motor driver to drive movements of robotic arm. Pick-and-place actions are performed by robotic arm, which precisely sorts objects with color into bins based on pre-defined color classes. Furthermore, an Internet of Things module is implemented to achieve remote monitoring and control and facilitate users to start, stop, and rewind processes remotely using a smartphone or web interface. The proposed system is suitable for installation in computerized shops, manufacturing industries, and intelligent waste disposal because it has greater object identification and sorting accuracy and provides remote access. The system's efficiency and real-time flexibility is validated by experimental results.

This paper describes how to use Cisco technology to create Software-Defined Wide Area Networking (SD-WAN). The study report focuses on how using SD-WAN technology may make operating enterprise networks easier and more efficient as demands alter. The main goal is to evaluate the functioning of the new infrastructure in terms of performance, scalability, and contrast with the existing WAN architecture, which includes Multiprotocol Label Switching. As part of the review process, the user experience was assessed in addition to network speed and latency. It was very clear from our evaluation that deploying SD-WAN improved agility and reduced costs. The results show a significant improvement in network performance, resource efficiency, and simple management. The importance of Cisco SD-WAN solutions in enhancing network performance and assisting companies in maintaining their competitiveness in the ever-changing networking landscape is demonstrated by this study.

The most common cause of dementia globally is cerebrovascular disease (CVD). There is yet no perfect method for identifying patients with cardiovascular disease who have vascular-cognitive-impairment (VCI). Neuroimaging and clinical non-imaging data from 421 individuals with CVD are gathered in this study. We used this information to build a multimodal deep-learning framework that used methods like vision transformer and extreme-gradient-boosting (EDB). The framework's final hybrid approach showed strong performance on both internal and external datasets using 2 neuroimaging characteristics and 6 clinical features. Additionally, our model was shown to have diagnostic performance that was analogous to that of expert doctors on a specific data set. Importantly, our model can pinpoint the specific areas of the brain and clinical characteristics that play a major role in the VCI diagnosis, making it easier to understand and use. This paper proposes a clinical decision support tool for identifying VCI in CVD patients that is both accurate and easy to understand. This research offers new understandings of how the kidneys age and indirectly supports clinical treatment decisions including dealing with kidney inflammation, stones, or tumors that may require nephrectomy, either partially or completely.

In this research a novel Privacy Oriented White Shark Encompassed hierarchical auto-associative polynomial Convolutional Neural NetwoRk (POWER) framework for task scheduling has been proposed. Initially, the Hierarchical Auto-associative Polynomial Convolutional Neural Network (HAP-CNN) for scheduling the healthcare task by considering the parameters. The HAP-CNN has been optimized using White Shark Optimization (WSO) for enhancing the accuracy in generating the schedule. The proposed task scheduling model is calculated based on several characteristics, including task migration, reaction time, transmission time, makespan, and cost. Recall, specificity, accuracy, precision, and F1 score were utilized to assess the proposed method's efficacy. With the suggested model, 99.32% classification accuracy was attained. The proposed model enhanced the total accuracy by 2.29%, 1.07% and 7.37% better than Task Scheduling utilizing a multi-objective grey wolf optimizer (TSMGWO), Prioritized Sorted Task-Based Allocation (PSTBA), and Large-Scale Industrial Internet of Things asynchronous Advantage Actor Critic system (LsiA3CS) respectively.

Efficient waste management is a pressing matter in an age of rising environmental concerns. The main purpose of this research is to design a proper waste classification system capable of distinguishing biodegradable waste from non-biodegradable waste. In addition, the suggested system employs advanced technologies and natural mechanisms to accurately classify garbage thus ensuring that right disposal methods are used during recycling initiatives. It serves as a means to separate compostable objects from those that cannot be broken down naturally and require specific ways of disposal making it possible for the environment to be sustained in its status quo. Hence, its implementation could weaken the effect that garbage has on the planet offering a purer and safer environment. Furthermore, further studies need to be conducted so as to improve the system for use in large-scale scenarios and wide range of waste types. This study is another significant step towards sustainable waste disposal and greener future.

In the current digital environment, copious amounts of data are generated across diverse sectors like healthcare, content creation, the internet, and businesses. ML algorithms are pivotal in analyzing this data to unveil significant ways to make decisions. However, not all features within these datasets are relevant for constructing robust machine learning models. Some features may be insignificant or have minimal impact on the prediction outcomes. By filtering out these irrelevant features, the computational burden on machine learning algorithms is reduced. Using the freely available MINIST dataset, this study explores the application of t-SNE, LDA, and Principal Component Analysis (PCA) alongside several prominent ML techniques like Naive Bayes, SVM classifiers, and K-NN classifications employed. Experimental outcomes illustrate the effectiveness of ML algorithms in this context. Furthermore, the experiments demonstrate that employing PCA with machine learning algorithms leads to improved outcomes, particularly when dealing with high-dimensional datasets. Performance measures like Accuracy 98.34%, Sensitivity 98.76%, Recall 98.45% and Throughput 98.65% have been attained, which was a good improvement.

All social networks allow image uploads and sharing. To allow users to exchange photographs, social networking sites use content-based recommendations (based on history), collaborative suggestions (based on the user and his friends' similarities), personalized advice, etc. Since no previous technique used socially advanced qualities like Upload History, Social Influence, or Owner Admiration, we can acquire a context relationship between people and images, which helps make optimal relationship-based suggestions. This generates a hierarchical attention model with three essential aspects and a CNN, where CNN represents the User's visual image model and three critical aspects reflect upload history, social influence, and owner matrix. The proposed application improved accuracy, sensitivity, and recall to 93.23%, 95.23%, and 97.73 %, respectively.

This paper illustrates a hybrid algorithm that is amalgamation of two bio-inspired algorithms such as Cuckoo search and Firefly optimization algorithms are incorporated to find optimize scaling factor intended for the watermark insertion. This combination is considered as one of the utmost usages in the recent past. The foremost benefit is obtained by combining several features of both the algorithms. SSIM and BER are used to evaluate fitness function in this optimization technique. Contemplating above opportunities, a novel video watermarking method is proposed by using CS-FA in DWT, SVD domain. Besides, the watermark security is also improvised using secret sharing method. Experimental tests reveal that the video watermarking approach suggested has a reasonable imperceptibility and an improved robustness against attacks.

Content-based image retrieval (CBIR) has become one of the trending areas of research in computer vision. In this paper, consonance on hue, saturation, and intensity is used by applying inter-channel voting between them. Diagonally symmetric pattern (DSP) from the intensity component of the image is computed. The grey level co-occurrence matrix (GLCM) is applied to DSP to extract texture features. Histogram of oriented gradients (HOG) features is used to extract the shape information. All three features are concatenated. To evaluate the efficiency of our method, five performance measures are calculated, i.e., average precision rate (APR), average recall rate (ARR), F-measure, average normalised modified retrieval rank (ANMRR) and total minimum retrieval epoch (TMRE). Corel-1K, Corel-5K, Corel-10K, VisTex, STex, and colour Brodatz are used. The experimental results show an improvement in 100% cases for Corel-1K dataset, 80% cases for Corel-5k and 80% cases for each of the three texture datasets.

The rapidly growing use of huge image database is becoming possible with the growth of multimedia technologies. Content-based image retrieval (CBIR) is observed as an efficient method for carrying out its management and retrieval. This paper embellishes the benefit of the image retrieval system based on the information as well as key technologies. Compared to the shortcoming that only one feature of the conventional method can be used, this paper proposes a technique for image retrieval, by analyzing a vigorous component descriptor named local majority intensity patterns (LMIP) for texture image retrieval. LMIP is the referenced pixel dependent on the encompassing lion’s share pixels’ conduct included in the image. The proposed LMIP have utilized the wager dominant part of odd and even pixels individually. The exploratory results have demonstrated that the proposed LMIP descriptor has accomplished a superior acknowledgment precision than existing methods by consuming less computation time.

The objective of this paper is to apply machine learning and visual processing to identify overworked IT employees. Our technology is an improved version of older stress detection systems that did not include live detection or personal counseling. Stress detection methods that don't include real-time monitoring or individual counselling are being updated in this research. A survey is used to collect data on employees' mental stress levels in order to provide effective stress management solutions. In order to get the most out of your employees, this paper will look at stress management and how to create a healthy, spontaneous work environment.

In this paper, processing of huge number of images is achieved to retrieve a queried image using MapReduce paradigm with different modes. These systems are useful in cases where the traditional single computer cannot process such huge image data. Nevertheless, such processing with a single computer system will take a long time to complete the processing. A total of six types of modes for processing the image data is proposed in this paper. To show the performance of the systems, the results are shown with different number of workers involved in processing the image data. The results show that the proposed MapReduce paradigm with different modes are performing as expected when there is a change in the number of workers involved in processing i.e., the time taken to complete the job is indirectly proportional to the number of workers considered. Even though the time to complete the task has changed, the performance measures: Precision, Recall, F-Measure, Retrieval Rank and Minimum Retrieval Epoch are same for all modes. The computational time for two image datasets: Corel1K and VisTex for a total of five image retrieval methods are evaluated. For completing all the five image retrieval methods on Corel1K, the time saved is 43%, 45% and 68% respectively for the number of workers as 4vs2, 2vs1 and 4vs1 workers. Similarly for VisTex it is 42%, 46% and 68%. The algorithm used for getting the features from the image are the authors recently published algorithms.

Development of Deep Learning models of Internet of Things (IoT) enclosures with limited resources are difficult because Both Quality of Results are difficult to achieve - QoR as follows two models, DNN Model, and Inference Accuracy and Quality of Services such as power consumption, throughput, and latency. Currently, the development of DNN models is often separated from deploying them to IoT devices, which leads to the most effective solution. If there are many records that represent objects of substantially the same class (face, human body, etc.), you can apply frames to each object of this class. To achieve this, use an independent representation to distinguish between appearance and progress data. Deep fake detection is achieved by using a novel, lightweight Deep Learning method on the IoT platform that is memory-efficient and lightweight. It is carried out in two different stages. The first phase of the deep fake test aims to implement a method of extracting images from a video and using them in conjunction with a Deep Neural Network to implement a test for face animation. It has been reported that the impact of the background elimination has been reported before the background subtraction. Here the Trans GAN model is used for the image classification. In the second phase, the work can be recorded and executed by the IOT device that can record live video streams and then detect activity involved in live video. An activity detection prototype based on IoT devices with small processing power is presented. This prototype provides improvements to the system, extending its application in various ways to improve portability, networking, and other equipment capabilities. The proposed architecture will be evaluated against four highly competitive object detection benchmarking tasks CIFAR10, CIFAR100, SVHN, and ImageNet.

Content-Based Image Retrieval (CBIR) has become one of the trending areas of research in computer vision. In traditional CBIR the features in spatial domain, such as color, texture, shape and point features are extracted. It is often considered that apart from the spatial features, the features extracted from the frequency domain of the images can give further information on the features of an image. This paper proposes two novel methods for the purpose of feature extraction from the 2-dimensional Discrete Cosine Transform (DCT) of an image. DCT-256-Zigzag and DCT-256-2×2. These methods take into considerations the lower frequencies in order to determine the features in the frequency domain. The advantage of using the zigzag scanning is to have the maximum low frequency values having Higher Energies comparatively. These two features are combined with two of the existing spatial domain features: Local Binary Patterns (LBP) and Interchannel voting features to generate a global feature vector for an image. For an query image, its feature vector is compared with feature vectors of every other image in the database using dl-distance and the images with least distance is considered most similar image to the query image. To evaluate the efficiency of these two methods, five standard performance measures such as Average Precision Rate (APR), Average Recall Rate (ARR), F-Measure, Average Normalized Modified Retrieval Rank (ANMRR) and Total Minimum Retrieval Epoch (TMRE) are used. Six benchmark image datasets: Core1-1000, Corel-5000, Core1-10000, VisTex, STex, and Color-Brodatz are used to corroborate the performance of these methods.

In this paper, we propose novel content-based image retrieval (CBIR) algorithms using Local Octa Patterns (LOtP), Local Hexadeca Patterns (LHdP) and Direction Encoded Local Binary Pattern (DELBP). LOtP and LHdP encode the relationship between center pixel and its neighbors based on the pixels' direction obtained by considering the horizontal, vertical and diagonal pixels for derivative calculations. In DELBP, direction of a referenced pixel is determined by considering every neighboring pixel for derivative calculations which results in 256 directions. For this resultant direction encoded image, we have obtained LBP which is considered as feature vector. The proposed method's performance is compared to that of Local Tetra Patterns (LTrP) using benchmark image databases viz., Corel 1000 (DB1) and Brodatz textures (DB2). Performance analysis shows that LOtP improves the average precision from 59.31% to 64.36% on DB1, and from 83.24% to 85.95% on DB2, LHdP improves it to 65.82% on DB1 and to 87.49% on DB2 and DELBP improves it to 60.35% on DB1 and to 86.12% on DB2 as compared to that of LTrP. Also, DELBP reduces the feature vector length by 66.62% as compared to that of LTrP. To reduce the retrieval time, the proposed algorithms are implemented on a Hadoop cluster consisting of 116 nodes and tested using Corel 10K (DB3), Mirflickr 100,000 (DB4) and ImageNet 511,380 (DB5) databases.

Research on Content-Based Image Retrieval is being done to improvise existing methods. Most of the techniques that were proposed use color and texture features independently. In this paper, to get the correspondence between color and texture, we use congruence on Hue, Saturation, and Intensity by using inter-channel voting. Gray Level Co-occurrence Matrix (GLCM) on Diagonally Symmetric Pattern is computed to get texture features of an image. The similarity metrics between two images is computed using congruence and GLCM. To measure the performance; Average Precision Rate (APR), Average Recall Rate (ARR), F-measure, Average Normalized Modified Retrieval Rank (ANMRR) are calculated. In addition to these parameters, one more parameter has been proposed: Total Minimum Retrieval Epoch (TMRE) to calculate the average number of images to be traversed for each query image to get all the images of that category. To validate the performance of the proposed method, it has been applied to six image databases: Corel-1 K, Corel-5 K, Corel-10 K, VisTex, STex, and Color Brodatz. The results of most of the databases show significant improvement.

A digital image can be represented by Weighted Finite Automata (WFA), where it uses Quadtree partitioning. An image of size 2n×2n pixels can be represented with WFA, which denotes the relation between different sub-parts (state images) of an image. Researchers have used WFA to represent an image in compressed form. We have proposed the method of identifying the regularity in a given texture by finding the number of state images generated from the textures. The method is tested on binary and gray textures. For color textures, the method is tested on RED, GREEN and BLUE components individually. The results clearly show the effectiveness of WFA in identifying the regularity in all the three types of regular textures.

Processing gigantic images with normal image processing techniques can be time consuming and difficult. Here gigantic mean with respect to dimension (Giga Pixels) or memory size (Giga Bytes). These images can be sometimes too large to load into the memory or they can be loaded, but then takes more time for processing. To overcome this problem, we proposed a method, Cluster Based Block Processing, to process large images by splitting the image according to their dimension or size and process it on different machine in the Hadoop cluster using Map-Reduce for effective processing. Representative results of comprehensive experiments on gigantic images are selected to validate the capacity of our proposed method over the traditional methods. Our results show that the proposed method is 20× faster than existing traditional methods.