Vijay Katkar

The challenge of identifying individuals in various situations, such as walking orientation and clothing conditions, from their gait patterns remains challenging in biometric recognition. To mitigate this challenge, this work presents a unified mathematical formulation that incorporates optical flow-based viewing angle estimation to reduce computational complexity and a Kernel-PCA (KPCA) ensemble framework, combining the strengths of multiple pre-trained Convolutional Neural Networks (CNNs), including EfficientB0 and B1, MobileNet and MobileNetV3 variants, as well as several RegNetX and RegNetY models. The most suitable lightweight model among pretrained CNNs is used to extract the features based on the estimated angle and Gait Energy Image (GEI) generated from the CASIA B dataset. The extracted features are refined using KPCA to improve class separability and recognition. These transformed features are utilized to train Machine Learning (ML) classifiers such as K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Random Forest, and XGBoost. From these classifiers, KNN using transformed features produced the best result. The proposed method achieved recognition accuracies between 95.6% and 99.2% across viewing angles from 0° to 180° in the training and 92.07% to 95.57% in testing. By integrating angle estimation with selective model usage, the framework significantly reduces computational overhead while maintaining high recognition performance. Collectively, the proposed approach delivers a robust and efficient solution for GEI-based person identification under a varied range of real-world conditions and viewing angles.

Effective detection of corn leaf diseases is vital for preventing significant agricultural losses. This paper presents a novel methodology that integrates Transfer Learning, Kernel Principal Component Analysis (Kernel PCA)-based feature enhancement, and bio-inspired optimization to improve disease classification on corn leaves. Several pretrained models, including MobileNet, VGG16, and various DenseNet and ResNet architectures, were employed to extract deep features from corn leaf images. Kernel PCA was used to capture nonlinear patterns, enhancing the models’ discriminative power. Classification was performed using the XGBoost algorithm, optimized through the Slime Mould Algorithm (SMA) for feature selection and hyperparameter tuning. Experimental results demonstrate that MobileNet achieved the highest performance with an accuracy of 99.08% and specificity of 99.39%, while DenseNet169 showed comparable effectiveness. The interpretability of the proposed framework was validated using Local Interpretable Model-Agnostic Explanations (LIME), which visually highlighted the critical regions of leaf images influencing disease classification decisions. This integrated approach provides a robust, accurate, and interpretable system for corn leaf disease detection.

Brain tumors are extremely dangerous and can lead to a drastically shortened life expectancy, making them an extremely hazardous illness. In order to create treatment programs that can prolong the lives of people impacted by brain tumors, a correct diagnosis is essential. It is difficult and time-consuming to manually identify and analyze massive amounts of MRI data. To successfully categorize brain tumors, we present a new DL method based on transfer learning in this research. We have conceptualized a novel strategy that incorporates fine-tuning, transfer learning architecture rebuilding, and thorough pre-processing. In our transfer learning implementation, we make use of Xception, VGG19, and MobileNetV3Small, among others. Xception achieved an accuracy score of 96.84%, whereas VGG19 achieved an accuracy score of 84.98%, using the public domain Kaggle MRI Brain tumor dataset that consisted of 253 pictures. Based on our research, Xception outperforms other models on the MRI Brain tumor dataset, reaching an impressive accuracy rate of 96.84%. To help neurologists and physicians make quick and accurate diagnoses for patients with brain tumors, our suggested model can effectively categorize tumors in a short amount of time.

In past several decades, gait biometrics has emerged as a viable alternative to traditional identification methods, offering advancements in surveillance, monitoring, and analysis techniques. However, determining gender based on gait remains a challenge, particularly in computer vision applications. This study proposes a robust and adaptable approach to address this issue by leveraging gait analysis. There is a growing need for datasets tailored to gait analysis and recognition to facilitate the extraction of relevant data. While most existing research relies on image-based gait datasets, this study utilizes the OULP-Age dataset from OU-ISIR, representing gait through gait energy images (GEIs). The methodology involves feature extraction from GEIs using pre-trained models, followed by classification with the XGBoost classifier. Gender prediction is enhanced through parameter fine-tuning of the XGBoost classifier. Comparative analysis of 11 pre-trained models for feature extraction reveals that DenseNet models, combined with optimized XGBoost parameters, demonstrate promising results for gender prediction. This study contributes to advancing gender prediction based on gait analysis and underscores the efficacy of integrating deep learning models with traditional classifiers for improved accuracy and reliability.

Machine learning is the latest approach to optimize the performance of absorbers, sensors, etc. A sensor with behavior prediction using polynomial regression is presented. Three different variations of metasurfaces namely double split-ring resonator, single split ring resonator, split ring resonator with thin wire are analyzed. The proposed design aims to achieve the highest sensitivity by observing different designs and different parameter variation. The highest sensitivity is achieved for double split-ring resonator and single split ring resonator designs. The change in thickness of different parameter affect the absorption and the highest sensitivity is calculated based on these variations. The polynomial regression (PR) model is employed to predict the absorption values for assorted combinations of intermediate wavelength values with angle variation, substrate thickness, substrate length, substrate width, graphene potential, and resonator thickness values. Test Cases R-30 and R-50 are evaluated using R² score metric to assess the effectiveness of PR model for predicting the values of absorption. R² score close to 1.0 is achieved for all the experiments at a higher (more than 5) polynomial degree, which proves the prediction efficiency of a regression model. The proposed biosensor designed with a PR model can be applied in biomedical applications for hemoglobin detection.

A highly efficient psi‐shaped solar energy absorber in the paper is developed. This structure is designed using titanium‐based psi shaped resonator, SiO₂ based substrate, and Tungsten metal‐based base layer to achieve a near‐perfect ultrawideband absorption spectrum under solar radiation. The average absorption of 97.04% is achieved in the observed range including the ultraviolet to mid‐infrared regimes and the ultrawideband characteristics are also achieved. It is also noted that for the bandwidth of 3730 and 2500 nm, above 90% and 95% absorption rate are achieved, respectively. The effect of several design parameters on the spectrum of absorption is explored and accordingly, optimized design is identified. Furthermore, the absorption response of the developed solar energy absorber is polarization‐insensitive and wide‐angle from 0° to 50°. Experiments are meant to identify the optimal parameter values for solar absorber design by employing a random restart hill climbing optimization approach. According to the experimental data, this strategy can reduce the computing power and time requirements of the simulation process by 97.57%.

Rapid detection of mycobacterium tuberculosis bacteria is very important in reducing tuberculosis disease. We propose a label-free graphene-based refractive index sensor using a machine learning approach that detects mycobacterium tuberculosis bacteria. The biosensor is designed for higher sensitivity by analyzing different parameters like substrate thickness, resonator thickness, and angle of incidence. Machine learning is applied to predict the values of absorption for different wavelengths. The machine learning model is applied to four different parameters (angle of incidence, substrate thickness, resonator thickness, graphene chemical potential) of the biosensor. The plus shape metasurface is placed above the graphene-SiO2 hybrid layer to improve the sensitivity. The comparative analysis with other published designs is also presented. The proposed sensor with its higher sensitivity and ability to detect mycobacterium tuberculosis bacteria can be used in biomedical devices for diagnostic applications. Experiments are performed to check the K-Nearest Neighbors (KNN)-regressor model’s prediction efficiency for predicting absorption values of intermediate wavelengths. Different values of K and two test cases; R-50, U-50 are used to test the regressor models using the R2 Score as an evaluation metric. It is observed from the experimental results that, high prediction efficiency can be achieved using lower values of K in the KNN-Regressor model.