2024
Permanent URI for this collectionhttps://dl.bhu.ac.in/bhuir/handle/123456789/36736
Browse
Search Results
Now showing 1 - 10 of 59
PublicationConference Paper A Cascaded Deep Learning Approach for Detection and Localization of Crop-Weeds in RGB Images(Springer Science and Business Media Deutschland GmbH, 2024) Rohit Agrawal; Jyoti Singh KirarWeeds compete with crops in the fields, thus lowering crop yield with losses of up to 80%. The efficient use of chemical herbicides is desired to reduce the harmful effects on the environment, which requires the location of the weeds to be known. In this paper, we present a deep learning approach capable of detecting and localizing weeds in RGB images, trained using the publicly available Open Sprayer dataset. The adopted methodology consists of a classification step using a pre-trained 2D convolution neural network and a Random Forest classifier, which is used to predict the presence of weeds in an RGB image. If presence is predicted, then an attempt to localize them has been done by cascading a segmentation step using a U-Net architecture. The proposed architecture can classify the presence of weeds in an image with an accuracy of 91.19% and predict the location of weeds in the image by generating binary masks with a mean Dice score of 0.879 on the publicly available Open Sprayer dataset. © 2024, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.PublicationConference Paper Principal Causes of Soil Erosion in a Watershed from the Ganga Basin, India: Evidence from Land Use Land Cover Dynamics(Springer Science and Business Media Deutschland GmbH, 2024) Nikita Shivhare Mitra; Akansha Rupal Nath; Khushboo Pachori; Shyam Bihari Dwivedi; Prabhat Kumar Singh DikshitSoil degradation is the primary issue faced by some of the countries including India. The main reason behind this problem is the soil erosion. So the primary objectives of this paper were to assess the potential impact of Land use land cover (LULC) dynamics and other principal causes of soil erosion on sediment yield. The evaluation of the impact of LULC dynamics was done with three scenarios (LULC of 2004 with climate data of 1995–2004, LULC of 2015 with climate data of 2005–2015, LULC of 2004 with climate data of 2005–2015). The soil and water assessment tool (SWAT) was used for all three scenarios to estimate the sediment yield. The calibration and validation of the SWAT model were done using the Sufi-2 algorithm. Multivariate linear regression technique was used to find out most dominating causes accountable for soil erosion. This study shows that due to the transition of Agriculture and Forest land to urban land and range land the average annual runoff is increased by 13% whereas the sediment yield is decreased by 26%. As the result of regression technique, we estimated that the coefficients of the variables for slope >40, Barren land type, agriculture land type and Soil type 3 (having Hydrological soil group C, with sandy Clay texture, and percentage of sand is dominant in this soil type) are the highest with values 2.95, 0.49, 0.25, and 0.13 respectively, which concludes that the slope is the dominant cause of soil erosion. Barren land and agricultural land are the most soil erosion prone LULC classes. Moreover, among the soil classes, soil type 3 was dominant causes of soil erosion. The R coefficient of the techniques was equal to 0.93 which shows the efficiency of the result. These results can be further utilized in land use planning and for applying effective measures for soil conservation. © 2024, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.PublicationConference Paper Some Observations on Social Media Mining tools for Health Applications(Springer Science and Business Media Deutschland GmbH, 2024) Ankita; Rakhi GargThe use of social media has immensely increased globally, which accordingly have generated tremendous amounts of data that have in turn attracted many researchers. Accessing, analyzing and mining this Big Social Data bring about a great deal of challenges. Social Media has ample data for various applications including the health domain that we have seen recently during the outbreak of COVID-19. Various applications and frameworks were designed for real-time monitoring and detection of disease spread, patient’s health status, providing information about active cases, hotspot areas, vaccination centers and also for disseminating government advisories. Social Media Mining provides tools and techniques to mine Social Media for extracting information, study patterns and analyze data. Social Media Mining can be used to bridge the gap between current technologies and healthcare systems. This paper basically reviewed various tools, techniques and algorithms developed for Social Media Mining for Health Applications (SMM4HA) and also discussed the major challenges that emanate using SMM4HA such as the issue of health misinformation and rumors, privacy and security issues, issue of data breaches, Big Social Data and the ethical challenges. We have also discussed some recent works in the healthcare domain such as disease surveillance, drug detection, disease prediction, etc., focusing on issues and challenges in machine learning–Natural Language Processing and use of medical ontologies as Social Media Mining Tools and Techniques that will help researchers and scientists working in this area. More research in the area is needed to be done considering rapid change of voluminous Social Media data which require efficient computation models as well as algorithms. © 2024, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.PublicationConference Paper Semantic Segmentation Based Image Signature Generation for CBIR(Springer Science and Business Media Deutschland GmbH, 2024) Suneel Kumar; Mona Singh; Ruchilekha; Manoj Kumar SinghContent-Based Image Retrieval (CBIR) leveraging semantic segmentation integrates semantic understanding with image retrieval, enabling users to search for images based on specific objects or regions within them. This paper presents a methodology for constructing image signatures, a pivotal element in enhancing image representation within a CBIR system. The efficiency and effectiveness of a CBIR system significantly hinge on the quality of the image signature, which serves as a compact and informative representation of raw image data. Our proposed methodology begins with emphasizing clear object or region boundaries through pixel-level semantic segmentation masks. A pretrained semantic segmentation model, such as DeepLab v3+, is employed to generate pixel-wise object class predictions, yielding the necessary segmentation masks. Subsequently, each image is segmented into meaningful regions based on these masks, and relevant features are extracted from each segmented region using a pre-trained Deep Convolutional Neural Network (DCNN) models AlexNet, VGG16 and ResNet-18. During the retrieval phase, when a user queries the system with an image, the query image is segmented using the pre-trained semantic segmentation model, and features are extracted from the segmented regions of the query image. These query features are utilized to search the database for the most similar regions or images. Similarity scores, calculated using Euclidean distance, are used to rank the database entries based on their similarity to the query, allowing for efficient retrieval of the top-k most similar regions or images. We found that for some classes semantic segmented based retrieval better performance in comparison to image based. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.PublicationConference Paper HUCMD: Hindi Utterance Corpus for Mental Disorders(Springer Science and Business Media Deutschland GmbH, 2024) Shaurya Prakash; Manoj Kumar Singh; Uma Shanker Tiwary; Mona SrivastavaAs our knowledge, there is no dialog system for mental health-care domain in Hindi. This may be due to unavailability of user utterances corpora in Hindi for this domain. In this paper, we propose a novel algorithmic approach for user utterance generation in Hindi by considering dialects, linguistic attributes, symptoms, frequency of symptoms, and intensity of symptoms and history of symptoms. We use nine symptoms (anger, emptiness, fear, irritation, restlessness, suicide, sadness, tension, worry) as given in DSM5, ICD-11, and WHO guideline. These symptoms were used for generation of utterances and validation of the generated utterances for different type of mental diseases. We collected utterances by interviewing patients in clinic and found that it closely match to the utterance generated by proposed algorithm. The generated utterance corpus is also validated using machine learning methods in the framework of CNN, Bi-LSTM and Dense. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.PublicationConference Paper Effect of Multidirectional Forging on Mechanical and Tribological Behavior of Hypereutectic Al–20wt.% Si Alloy(Springer Science and Business Media Deutschland GmbH, 2024) Ishwari Narain Choudhary; Nitesh Kumar Sinha; Manik Mahali; Jayant Kumar SinghEnhancement in the material properties for a specific application is of major concern in all types of industries, e.g., automobile, aerospace, marine, etc., grain size plays an important role in affecting the mechanical and tribological properties of metals and alloys. The present work investigates how the hyper-eutectic Al–Si alloy properties can be affected by the application of multidirectional forging (MDF). Hyper-eutectic Al–Si alloy containing 20%Si was prepared by casting and a 40% reduction in height in each direction at 300 ± 10 ℃ was given using a power hammer machine during forging. The results show that mechanical and tribological properties such as tensile, hardness, density, porosity, and wear properties improved. The microstructural improvement from dendrites to fine grain size also happened and was uniformly distributed after multidirectional forging. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024.PublicationConference Paper GenEmo-Net: Generalizable Emotion Recognition Using Brain Functional Connections Based Neural Network(Springer Science and Business Media Deutschland GmbH, 2024) Varad Srivastava; Ruchilekha; Manoj Kumar SinghThe aim of this research is to construct a generalizable and biologically-interpretable emotion recognition model utilizing complex electroencephalogram (EEG) signals for realizing emotional state of human brain. In this paper, the spatial-temporal information of EEG signals is used to extract brain connectivity-based feature, i.e., phase-locking value (PLV), that incorporates phase information between a pair of signals. These functional features are then fed as input to our proposed model (GenEmo-Net), which encompasses of Graph Convolutional Neural Network (GCNN) and Long Short-Term Memory Network (LSTM). It is able to dynamically learn the adjacency matrix that resembles functional connections in the brain, and are combined with the temporal features learnt by LSTM. To validate the generalization ability of our model, the experimental setup combines three emotion databases, namely DEAP, DREAMER, and AMIGOS, which increases variability and reduces biasness among subjects and trials. We evaluated the performance of our proposed model on the combined dataset, which achieved a classification accuracy of 70.98 ± 0.73, 65.47 ± 0.56, and 70.09 ± 0.37 for discrimination of valence, arousal, and dominance, respectively. Notably, our generalized model gives more robust results for emotion recognition tasks when compared to other methods. In addition, the biological interpretation of GenEmo-Net is tested via the final adjacency matrix, learnt at the end of training, for VAD processing units. Above results demonstrate the efficacy of the GenEmo-Net for recognizing human emotions and also highlight substantial variations in the spatial and temporal brain characteristics across distinct emotional states. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.PublicationConference Paper Optimizing Resource Utilization and QoS-Conscious Application Deployment Through AHP in Edge Computing(Institute of Electrical and Electronics Engineers Inc., 2024) Yasasvitha Koganti; Ram Narayan Yadav; Ajay PratapEdge computing has emerged as a promising technology to satisfy the demand for data computational resources in Internet of Things (IoT) networks. With edge computing, processing of the massive data-intensive tasks can be in the proximity of IoT users. Thus, the required constraints related to tasks such as resource requirements, and the quality of service can be guaranteed. However, the question that how to determine the task offloading strategy under various constraints of resources, distance, and cost is still an open issue. In this paper, we study the task offloading problem from a matching perspective and propose an Edge-User Assignment Algorithm (EUAA) that aims to maximize the resource utilization of edge servers while satisfying the Quality of Service (QoS) requirement of IoT users. In any matching algorithm, the main concern is how to generate the preference order of either side. To generate preference orders for edge servers, we use the concept of the Analytical Hierarchy Process (AHP). We have considered the following criteria: distance from users to the server, latency, available resources, and pricing. This generates the priority of the users for matching to edge servers. From IoT users' perspective, we use cost, and QoS parameters to improve their satisfaction. We compare the performance using the number of assigned users, servers' profit, number of satisfied users, and edge server resources used. The simulation results confirm that significant profit and satisfied users can be achieved by the proposed algorithm. © 2024 IEEE.PublicationConference Paper Emotion Recognition Using Phase-Locking-Value Based Functional Brain Connections Within-Hemisphere and Cross-Hemisphere(Springer Science and Business Media Deutschland GmbH, 2024) Ruchilekha; Varad Srivastava; Manoj Kumar SinghResearch in cognitive neuroscience has found emotion-induced distinct cognitive variances between the left and right hemispheres of the brain. In this work, we follow up on this idea by using Phase-Locking Value (PLV) to investigate the EEG based hemispherical brain connections for emotion recognition task. Here, PLV features are extracted for two scenarios: Within-hemisphere and Cross-hemisphere, which are further selected using maximum relevance-minimum redundancy (mRmR) and chi-square test mechanisms. By making use of machine learning (ML) classifiers, we have evaluated the results for dimensional model of emotions through making binary classification on valence, arousal and dominance scales, across four frequency bands (theta, alpha, beta and gamma). We achieved the highest accuracies for gamma band when assessed with mRmR feature selection. KNN classifier is most effective among other ML classifiers at this task, and achieves the best accuracy of 79.4%, 79.6%, and 79.1% in case of cross-hemisphere PLVs for valence, arousal, and dominance respectively. Additionally, we find that cross-hemispherical connections are better at predictions on emotion recognition than within-hemispherical ones, albeit only slightly. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.PublicationConference Paper Machine Learning Strategies for Analyzing Road Traffic Accident(Springer Science and Business Media Deutschland GmbH, 2024) Sumit Gupta; Awadhesh KumarRoad safety and accidents have been an important concern for the entire world and everyone is putting effort into resolving the long-standing problem of road safety and accidents. In every country on earth, there is traffic and reckless driving. This has a negative impact on a lot of pedestrians. They become victims, although having done nothing wrong. The number of traffic accidents is rising quickly due to the enormous increase in road cars. Accidents like these result in harm, impairment, and occasionally even fatalities. Numerous things like weather changes, sharp curves, and human error all contribute to the high number of traffic accidents. In this research paper various machine learning techniques such as, K Nearest Neighbors, Random Forest, Logistic Regression, Decision Tree, and XGBoost etc., are used to investigate why road traffic accidents occur in various nations throughout the world. For evaluating and analyzing these algorithm several metrics, including precision, recall, accuracy and F1-Score are used to improve the performance of the dataset and predicts accuracy by approximately more than 85%. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.