Advancing plant disease classification using an attention-based CNN for intra-dataset and cross- dataset training Prateek Mahapatra, Madhumita Panda, Santanu Kumar Dash, Umesh Kumar Sahu Scientific Reports, 2026 The precise classification of plant diseases is crucial for ensuring food security for all people and boosting agricultural productivity. Although there has been significant progress in this field using deep learning approaches, cross-dataset training hasn’t drawn as much attention from researchers as intra-dataset training has. Moreover, very few models have successfully blended intra-dataset and cross-dataset training approaches. This paper proposes a novel attention-based Convolutional Neural Network (CNN) to overcome these limitations. The model improves feature extraction and classification accuracy across multiple datasets by using attention mechanisms. It was tested on five datasets (Digipathos, Northern Leaf Blight (NLB), PlantVillage, PlantDoc, and the CD&S dataset) that covered leaf diseases of both corn and potatoes. During intra-dataset training, the model achieved the highest classification accuracy of 99.38% when trained on images of potato leaves from the PlantVillage dataset. During cross-dataset training, the model exhibited the highest average classification accuracy of 82.93% for corn leaf diseases when trained on images from the CD&S dataset with their backgrounds removed. When compared to the techniques taken into consideration in this study under comparable experimental conditions, the results demonstrate improved performance. This study shows how the model may be flexible for both intra- and cross-datasets, offering a flexible way to categorize diseases that affect plants. Because of its ability to generalize across different datasets, it may be helpful in real-world agricultural applications with a wide variety of image quality and situations. This encourages the advancement of precision farming techniques and disease control.
Autonomous object tracking with vision based control using a 2DOF robotic arm Umesh Kumar Sahu, Mebin K. S., Abhinav K., Muhammed Muzammil P, Ankur Jaiswal, et al. Scientific Reports, 2025 The tracking of moving object by implementing robot manipulator is one of the challenging task for many applications such as manufacturing, agriculture, logistics, healthcare, space, military, entertainment, etc. In the deployment of robotic manipulators with real-time object tracking for aforementioned important applications, the proper sensor surveillance and ensuring stability are major challenges. The purpose of this study is to design a precise and responsive object-tracking system by eliminating the complexities related to tedious mechanisms, rigidity, requirement of multiple sensors, etc. which are commonly associated with traditional systems. The robotic arms can be effectively designed to track moving objects autonomously with vision-based control. In comparison with different classical and traditional servoing approaches, the image-based visual servoing (IBVS) is more advantageous in vision-based control. The present article describes a new approach for IBVS-based tracking control of 2-degree-of-freedom (DOF) robotic arm by including object identification and trajectory tracking based crucial components. To solve the issues associated with IBVS, an accurate deep learning-based object detection framework is employed. The presented framework is utilized to detect and locate the objects in real-time. Further, an effective vision-based control technique is designed to control the 2-DOF robotic arm with the help of real-time response of object detection system. The validation of proposed control strategy is done by performing a simulation and experimental investigations with CoppeliaSim robot simulator and 2-DOF robotic arm, respectively. The findings reveal that the proposed deep learning controller for the vision-based 2-DOF robotic arm achieves good levels of accuracy and response time while performing visual servoing tasks. Furthermore, thorough discussion on possibility of using data-driven learning technique has been explored to improve the robustness and adaptability of the presented control scheme.
Energy management with control parameter optimization for a PV/FC/battery islanded DC microgrid Harin M Mohan, Santanu Kumar Dash Engineering Research Express, 2025 The integration of renewable energy resources (RERs) and the need for reliable and efficient power distribution in emerging power systems has driven growing interest in microgrids. This has led to challenges like uneven load distribution, overloading, less usage of power, and unstable operations that may lead to damage to power electronic devices connected. The existing metaheuristic-based control mechanisms struggle to manage fluctuating sources and demands effectively, often leading to suboptimal energy management. Therefore, with such constraints, this research work proposes a method utilizing Sparrow Search Optimization (SSA) and optimizes the control parameters for more effective EMS functionality. To support efficient power management and balancing of generation versus load demands, the proposed system integrates dispatchable distributed generators such as fuel cells (FC) and battery storage systems (BSS) with non-dispatchable distributed generators such as solar photovoltaic (PV) source. The method uses SSA optimization of controller parameter estimation ensuring better stability and reliability, compared to conventional PI and other metaheuristic methods. The intended system was tested under various operating scenarios with the use of MATLAB Simulink, to validate the efficacy of this system in the regulation of renewable energy fluctuations and optimization of battery performance. The performance of the developed system was compared with systems optimized through Particle Swarm Optimization (PSO), Grey Wolf Optimization (GWO) and Harris Hawk Optimization (HHA). Simulation results show that the system optimized using the SSA method results in better efficiency with faster execution time than other optimization-based systems.
Meta model approach for real time and short-term forecasting of wind turbine power generation with interactive 24 h dashboard Gaurav Chauhan, Sachin Jakhar, Priyamvad Singh, Santanu Kumar Dash, Muchenedi Hari Kishor Engineering Research Express, 2025 This study evaluates the performance of various machine learning algorithms (Linear Regression, SVR, AdaBoost, XGBoost, Gradient Boosting, Decision Tree, Random Forest, Extra Trees, CatBoost) for predicting wind power generation. We investigate their strengths and weaknesses through extensive experimentation using data from Kaggle and ENTSO-E. To enhance accuracy, we employ a meta-model approach and incorporate data cleaning techniques. We integrate statistical methods, artificial neural networks, and deep learning for improved short-term forecasting. A key outcome is the development of a real-time GUI dashboard that utilizes the OpenWeather API to fetch wind data and display predictions. This user-friendly interface features visualizations, alerts, and real-time data updates. Our results demonstrate that the selected meta-model significantly surpasses traditional methods, achieving superior metrics like R-squared and RMSE. This research showcases the potential of hyperparameter-tuned machine learning for precise wind power prediction, contributing to increased renewable energy utilization and reduced greenhouse gas emissions.
Performance Analysis of MAF-Based Control Strategy in Hybrid Renewable Energy System Mohammed Shijas V, Saichol Chudjuarjeen, Santanu Kumar Dash, Muhammed Inamu Rahman Thekkedath, Harin M Mohan, et al. Proceedings of the 2025 3rd International Conference on Cyber Physical Systems Power Electronics and Electric Vehicles Icpeev 2025, 2025
