Journal of Computer Science Engineering and Software Testing

Explainable Hate Meme Detection Using Multimodal Learning

2026-02-17T07:00:04+00:00

Memes have become a popular form of communication on social media, combining images and text to express opinions, humor, and emotions. While many memes are harmless, some are used to spread hate, discrimination, and offensive stereotypes. Detecting such hate memes is challenging because the meaning often depends on both visual content and textual context, making traditional text-based approaches insufficient. This paper presents an Explainable Hate Meme Detection System (EHMDS) that analyses memes using a multimodal approach. The proposed system processes textual information using transformer-based language models and visual information using deep learning-based image encoders. These features are combined through a cross-modal attention mechanism to identify hateful content more effectively. In addition to classification, the system provides explanations by highlighting important words and image regions that contribute to the final decision.

Experiments conducted on the Facebook Hateful Memes Dataset demonstrate that the proposed system can accurately detect hateful memes while also improving transparency and interpretability. By providing human-readable explanations alongside predictions, the system supports more trustworthy, ethical automated content moderation.

Literature Survey Paper: GS-GA-SVM - Hybrid Grid Search-Genetic Algorithm for SVM Parameter Optimization in Medical Diagnosis

2026-04-07T09:13:58+00:00

Support Vector Machines (SVMs) are among the most powerful and widely used classifiers in medical diagnosis, but their performance is critically dependent on the proper selection of hyperparameters. Traditional parameter optimization methods such as grid search are computationally expensive and suffer from discretization errors, while genetic algorithms may converge prematurely and require extensive tuning. Hybrid approaches that combine the systematic exploration of grid search with the adaptive search capability of genetic algorithms have emerged as a promising solution. This literature survey paper provides a comprehensive review of hybrid Grid Search-Genetic Algorithm (GS-GA) frameworks for SVM parameter optimization in medical diagnosis. The paper examines the theoretical foundations of SVM hyperparameters (C, γ), the limitations of standalone optimization methods, and the synergistic advantages of hybrid approaches. It critically analyzes sequential two-stage architectures where coarse grid search identifies promising regions followed by GA fine-tuning, as well as integrated approaches where GA is enhanced with grid-based initialization and adaptive operators. The survey evaluates empirical results from comparative studies, demonstrating that hybrid GS-GA approaches achieve 1-2% improvement in classification accuracy while reducing computational time by 60-80% compared to exhaustive grid search. Key findings reveal that GS-GA hybrids effectively balance exploration and exploitation, escape local optima, and produce more stable parameter sets across different data splits. The paper also identifies research gaps, including the need for adaptive threshold selection, parallel implementations for large-scale data, integration with feature selection, and validation on multi-class medical problems. Furthermore, it explores recent advancements in hybrid optimization, including combinations with particle swarm optimization, Bayesian optimization, and chemical reaction optimization. The survey concludes that GS-GA-SVM represents a powerful and efficient approach for developing high-performance diagnostic systems with minimal manual tuning, offering significant potential for clinical deployment where both accuracy and computational efficiency are paramount.

PRAGFIX: A Firebase Powered Intelligent Campus Governance Platform for Sustainable Issue Management

2026-03-12T04:39:32+00:00

PRAGFIX is a digital platform designed to improve how universities and colleges manage campus infrastructure issues. Traditionally, infrastructure problems are reported in person, recorded in logbooks, or communicated through scattered channels, making it difficult to track responsibilities, monitor progress, and ensure timely resolution. PRAGFIX addresses these challenges by providing a centralized, real-time system that simplifies reporting and resolution of campus-related problems. The platform enables students, staff, and administrators to report issues efficiently, with each user group having role-based access. By offering real-time updates, PRAGFIX ensures transparency and helps stakeholders understand the severity of issues, their departmental allocation, and overall campus status. This structured approach enhances accountability and speeds up decision-making. PRAGFIX is built using Firebase technologies, including Firestore, Cloud Functions, and Firebase Authentication. These tools ensure scalability, secure access, and real-time data synchronization, allowing the system to handle multiple users simultaneously while maintaining data security. Additionally, PRAGFIX incorporates a “Green Points” system to encourage environmental responsibility and sustainable practices within the campus community. Overall, PRAGFIX promotes collaboration among students, staff, and administrators, supports informed decision-making, and contributes to creating a more efficient, transparent, and sustainable campus environment.

Early-Stage Qualifier and Disease Classification for Pomegranate using Deep Learning

2026-04-07T12:05:19+00:00

Pomegranate cultivation plays a vital role in India’s agricultural economy. However, diseases such as Bacterial Blight, Alternaria, Anthracnose, and Cercospora severely affect crop yield and fruit quality, leading to significant economic losses for farmers. This paper presents an intelligent, web-based system that automates pomegranate disease detection and stage prediction using a combination of deep learning and machine learning models. The proposed system employs MobileNetV2 and ResNet50 architectures for image-based disease classification and an SVM–Random Forest ensemble for predicting the severity stages (Low, Medium, High) using texture-based GLCM features. A custom dataset of healthy and diseased fruit images was used for training and validation, with preprocessing and augmentation to improve generalization. The web application integrates Flask (backend) and React (frontend) to provide real-time image uploads, camera-based detection, and dynamic multilingual advisories in English, Kannada, Telugu, Tamil, and Hindi. The ensemble model achieved a classification accuracy of 98.4% for disease identification and 84% for stage prediction, demonstrating robust performance and practical applicability. This research contributes toward precision agriculture by offering an accessible, AI-powered advisory tool that supports early detection, reduces crop loss, and empowers farmers with actionable insights in their local languages.

NASA Missions Dashboard: A Real-Time Space Analytics and Visualization System Using Streamlit and NASA APIs

2026-04-02T11:37:52+00:00

This paper details the creation, execution, and evaluation of a complete, real-time Space Analytics Dashboard that incorporates multiple NASA APIs with cutting-edge web technologies to create an interactive format for the illustration, visualization, and professional execution of data pertaining to space missions. It provides a comprehensive solution to the problem of disparate Space Mission Data by allowing users to access both historical mission data and live astronomical data (e.g., Astronomy Picture of the Day, NEO Tracking, International Space Station Live Feeds) through a single portal. The Dashboard’s User Interface is built using Streamlit, with SQLite as the Database Storage mechanism, and Plotly for creating interactive visualizations, and provides Key Performance Indicators (KPIs), Multi Dimensional Filtering Options, and Real Time Data Updates. The implementation successfully integrates different NASA Data Sources with API call response times of less than 3 seconds and Visualization Rendering Times of less than 1 second. The Dashboard achieved an API Reliability Percentage of 98.7% for APOD, 96.2% for NeoWS, and achieved Linear Scalability with 10,000 Historical Mission Records. User Evaluation was completed with 25 participants, who provided an average rating of 4.6 out of 5 for User Satisfaction, and significant improvements in User Ease of Use and Information Clarity compared to previous Products. The architecture of this system supports Automatic Data Refresh Mechanisms, Theme Customization, and Strong Error Handling Capabilities, making this product ideal for use in the areas of Education, Research, and Public Engagement for the field of Space Informatics.