DeepRice: A Deep Learning and Deep Feature-based Classification of Rice Leaf Disease Subtypes

+The study utilized CNN’s for early identification of Rice leaf diseases. The data collected consisted of 5932 images of rice leaves comprising 4 types of diseases & 900 healthy leaf images.

+Uniquely sub-classified into mild and severe, the images were augmented and 6 transfer models were trained for 50 epochs.

+A Custom CNN model was built with additional Dropout’s and Dense Layers achieving an improved accuracy of 99.94%.

Study On Mental Health and Obesity among Children Based on Daily Activities and Effects using Machine Learning

+The study investigated the correlation between daily activities and mental health in children aged 12 to 17. Focusing on those with mental health issues leading to stress, depression, and unhealthy eating habits, the study employs historical and statistical data, and utilized the extra trees algorithm to determine feature importance.

+The findings categorized daily activities based on their impact, offering insights into major and minor contributors to conditions like obesity.

+The study aims to inform future research on the intricate relationship between daily activities, mental health, and obesity, ultimately contributing to the long-term well-being of children as they transition into adulthood.

Chatbots Embracing Artificial Intelligence Solutions To Assist Institutions in Improving Student Interactions

+The study delved into the integration of Chatbot systems within the education sector, highlighting the constructive influence of Chatbot technology on students and institutional management through efficient service delivery. Employed Python frameworks such as Flask, HTML, CSS, and JavaScript for Front End Development.

+The research employed PyTorch, along with Natural Language Toolkit (NLTK) to categorize user messages and NLP techniques including tokenization and stemming to allow Chatbots perform semantic, sentiment analysis, so to provide periodic outcomes from a diverse range of potential responses.

Feature Engineering and Machine Learning for IoT-based Applications: An Overview of Algorithms

+The paper extensively reviewed machine learning algorithms for feature estimation, highlighting its strengths, weaknesses, and suitable applications. It addressed challenges in feature interpretation for IoT applications, proposing strategies like data preprocessing and model optimization to overcome issues such as data variability and resource constraints.

+A comparative analysis of algorithms based on accuracy, speed, and efficiency served as a practical guide for selecting the most appropriate algorithm for range-based prediction applications in IoT scenarios.

GAN-Based Facial Feature Reconstruction for Improved Masked Face Recognition during Covid

+We proposed a GAN-based approach to learn the underlying relationships between masked and unmasked facial features and generate plausible reconstructions of the missing facial features.

+The suggested strategy produced high-quality and precise reconstructions of the missing facial features, and it can be used for face recognition tasks.

+The findings suggested that GAN-based face reconstruction has the potential to overcome the limitations posed by face masks, providing a solution that preserves individual privacy and security while ensuring accurate face recognition.

Modeling Precipitation Patterns in Alappuzha-Kerala: An Analysis of Regression and Time Series Approaches

+The study brings out the relationship between meteorological parameters in Alappuzha, a district in Kerala, India. The analysis was made using the real-time data obtained from NASA's POWER project.

+The experiment considered T2M, RH2M, WS2M, SST, and PETCORR parameter values and employed statistical analysis techniques, including correlation, regression, and time series Modeling, to examine the direct relationship between variables and forecast future precipitation levels.

+Upon careful curation, visualizations provided us with better insights to identify trends within the data. A comprehensive comparison helped us identify the best model using mean squared (MSE) and r-squared (R2) values.

Multi-Document Summarization Made Easy: An Abstractive Query-Focused System Using Web Scraping and Transformer Models

+We proposed a web-based abstractive QF-MDS that simplified the process of summarizing multiple documents on any given topic. The system uses techniques, including web scraping, natural language processing, and transformer models, to automate the summarization process improving the ease of accessibility to information for users.

+The system was designed to take user input in the form of a query, the number of words to be summarized, and the number of documents to be referred to. It then utilizes Google search engine API integration to retrieve the most relevant webpages based on their ranking, and performs web scraping of tags using beautiful soup (bs4) and selenium frameworks.

+The scraped data undergoes pre-processing, including stop word removal, tokenization (using Auto tokenizer) , and visualizing frequency matrix and word-cloud plots (using seaborn and matplotlib). The transformer model ranks the words based on frequency and generates a summary of the text that is coherent, concise, and relevant to the user's query.

A Web-based Real-time Sales Data Analysis Using LSTM Model for Better Insight

+A machine learning-based web application was deployed in real-time for the company Nellai Agencies, to analyse their monthly sales data. An LSTM model was trained with data to predict the sales for the next 'n' days.

+The performance of the model was evaluated with the metrics such as Mean Absolute Error (MAE) & R2Score. The results obtained were used to make insights by the company by visualizing predicted graphs.

+Concluded that, the impact of using a machine learning-based web application supported better decision-making, including the distribution of sales & production, finding the prominent customer, increasing the profitability of the company.

Design and Evaluation of a Brain Signal-based Monitoring System for Differently-Abled People

+We presented a method to help medical professionals correctly infer therapies by using the six emotions that patients express. The classes included: Turning on the fan, Leisure, Draining, Starving, Need to Use the Toilet, and Time to Take Prescription.

+Following the analysis of the baseline features, machine learning algorithms were trained, their effectiveness was assessed, and the result was then input into a planned program to notify the caretaker over a Bluetooth headset.

+The voting classifier was determined to be the best model, since it demonstrated the highest likelihood of correctly predicting the category.

Prevention of School Shootings Using Neural Networks and Computer Vision

+The delay in response during voluntary gun shootout events has led to questioning the security in schools premises. To enhance security conditions, an automated system that utilizes DNNs to authenticate identity of students through facial recognition (HAAR Classifiers) was proposed.

+YOLOv3 (Object detection algorithm) was integrated to identify harmful weapons, ensuring safety, along with Colour identification algorithms to distinguish anomalies.

+The captured images of suspects on the live feed were automatically directed to authorities for quick inspection ensuring a quicker passage for communication, utilized IOT devices (buzzers) to raise alerts in the nearby stations.