Search

Convolutional neural networks (CNNs) and linear binary pattern algorithms that have been trained on specialised datasets provided by the Saveetha School of Engineering will be utilised in this project with the intention of developing and evaluating facial recognition systems that are capable of identifying emotions. Instructions & Materials Required: According to the findings of this research, lines of binary patterns and convolutional neural networks are two of the techniques that are utilised by the facial recognition system in order to determine the feelings that are displayed on the face. Through a series of iterations (N=10) through the database, the accuracy of the deep learning algorithms, which included CNNs and LBPs (Linear Binary Patterns), was increased. Ten photographs were taken from each of the forty students in the class, and GPower Software 3.1 was used to extract these photographs for each of the groups. A confidence interval of eighty percent, an alpha level of ninety-five percent, and a beta degree of two thousand two percent were all included in the statistical study. Following this, the four hundred images that were produced were separated into two distinct groups: Group 1 and Group 2. Both the openCv module, which employs a Convolutional Neural Network, and the outcomes of a face recognition system that identifies emotions through the use of a Linear Binary Pattern, which was created in Python, were compared and contrasted. When it comes to creating predictions, the convolutional neural network has an accuracy rate of 83.47 percent. Linear Binary Patterns have a higher degree of accuracy, coming in at 82.15 percent. Based on the results of the Independent sample T-test for comparing means, a significant value of 0.0405 (p<0.05) was achieved. This indicates that the newly developed method for identifying emotions accurately captures the feelings of the student. The significance value is a statistical measure that confirms the relevance of the two categories. Conclusions: Linear Binary Patterns and Convolutional Neural Networks are the two components that make up the Face Recognition system, which is supposed to determine the feelings that are communicated through the face. The results of this study indicate that Convolutional Neural Networks are superior to Linear Binary Patterns in terms of accuracy. The statistical significance of the relationship between the two categories is indicated by a p-value of 0.0405. [ABSTRACT FROM AUTHOR]

The agricultural sector has a great influence on the economic and social status of human beings. Unsustainable management can lead to agricultural losses. Disease affects the health of the plant, which will negatively affect its growth. It is crucial to monitor the growth of the cultivated crop to ensure minimal loss. There is a class of deep learning called Convolutional Neural Network that is used for classification of images, segmentation of images, and signal processing among other mainstream tasks. The aim is to improve the performance of traditional models of disease detection by using the simplest approach and utilizing minimal computer resources. DenseNet-121 is used for detecting and predicting plant diseases. Automatic feature extraction is used by neural network models to aid in classifying input images into disease classes. 98.9% accuracy has been achieved with the proposed neural network approach, indicating its feasibility even under unfavorable conditions. [ABSTRACT FROM AUTHOR]

Multi-Level Inverters (MLI) are the state of the art technology used for harmonic reduction in inverters. Selective Harmonic Elimination-Pulse Width Modulation (SHE-PWM) technique and Sinusoidal PWM (SPWM) technique are the preferred modulation technique for improvement of harmonic profile in Symmetric Cascaded Multi-Level Inverter (SCMLI). The objective function of SHE-PWM is a challenging task in multilevel inverter as the function is non-transcendental in nature with multiple local minima. Thus Bio-geographical Based Optimization (BBO) and Particle Swarm Optimization (PSO) algorithm are applied to optimize the switching angles of SHE-PWM technique. The simulated harmonic spectrum of seven-level SCMLI with SHE-PWM technique is compared with the SPWM technique. SHE-PWM technique has completely eliminated the specified lower order harmonics with the harmonic component h5 and h7 to less than 0.8%. SPWM technique had minimized the overall THD at the cost of increase in amplitude of lower order harmonics h7 to 2.%. SHE-PWM is superior in performance when compared to SPWM technique due to elimination of lower order harmonics at lower switching frequency. [ABSTRACT FROM AUTHOR]