Jimmy Dani

Jimmy Dani

PhD Student

Dept. of Computer Science & Engineering

🧑‍💻 Projects

Generative AI

🎥📄 Audio/Video Summary Generator using OpenAI’s Whisper API

An intelligent tool that automatically transcribes and summarizes video or audio content using OpenAI Whisper API. This project demonstrates how to extract meaningful insights from media files by combining audio processing, speech-to-text, and large language model summarization using Python.

🤖💬 Chatbot using OpenAI and Flask

A conversational AI chatbot built using OpenAI’s language models and Flask. This project demonstrates how to integrate Gen AI capabilities into web applications for interactive conversations.

Course Projects

🌊🗺️🔍 Analysis of Flooded Areas based on Image Processing

Floods are natural disasters that cause the most economic damage at the global level. Therefore, flood monitoring and damage estimation are very important for the population, authorities, and insurance companies. The paper proposes an original solution, based on a Local Binary Pattern (LBP), Gray-Level Co-Occurrence Matrix (GLCM) and Statistical Moments with K-means clustering solve to this problem. In this paper in order to evaluate flood damage, two tasks are accomplished: the area coverage by flood and segmentation techniques. Finally, the experimental results in a real mission demonstrate the validity of the proposed methodologies and the performances of the algorithms.

🐄🔍📊 Cattle Detection and Counting using YOLO (You Only Look Once)

Accurate livestock detection and enumeration are critical for precision agriculture and large-scale farm management. This work presents a robust cattle detection and counting framework built on the YOLO (You Only Look Once) object detection paradigm, optimized for aerial imagery acquired via unmanned aerial vehicles (UAVs). A high-quality training dataset was constructed by extracting cattle instances from UAV-captured videos, followed by a tailored preprocessing pipeline using the OpenCV library. The pipeline incorporates resizing, denoising, segmentation, and morphological smoothing to enhance salient features and suppress background noise. These processed samples were used to train a convolutional neural network (CNN) integrated within the YOLO framework, enabling high-precision detection under challenging conditions such as variable illumination, occlusions, and heterogeneous terrain. Experimental evaluation demonstrates that the proposed method achieves 90% detection accuracy with an average counting time of 0.30 seconds per frame, providing near real-time performance suitable for scalable livestock monitoring in operational farm environments.

🧬🧠 Evolutionary Neural Network for Predicting Oil Permeability using Soft Computing Techniques

Permeability is one of the important characteristics of an oil field, in fact, it is a key parameter in describing a hydrocarbon reservoir. Numerous specialists have dealt with penetrability estimation strategies, however, there is no all-inclusive technique yet which can anticipate porousness in the entire field and in all interims of the wells. So computerized reasoning strategies have been utilized to anticipate penetrability by utilizing admirably log information in all field regions. In our research, we are trying to use machine learning techniques to predict the permeability of the oil. Data from four oil wells were used and different dataset patterns were constructed to evaluate performances of the models in predicting permeability by using either previously seen data or unseen data. Results obtained by using these machine learning techniques have acceptable performance in the prediction permeability of the oil. Machine Learning Algorithm Multilayer perceptron is being used for predicting the permeability of the oil. Evolutionary algorithms, namely, genetic algorithm and particle swarm optimization are being used to tune the parameters of the MLP neural network. Our conclusion shows the particle swarm optimization performs relatively better than a genetic algorithm for training MLP neural network.

🔒🛡️💻 Secure and Privacy-Preserving Online Student Feedback and Lecturer Evaluation System

Secure collection and analysis of student feedback are critical for ensuring the integrity of academic evaluation processes. This work presents a secure, web-based system for lecturer evaluation that enables authenticated students to submit feedback while preserving privacy and preventing unauthorized access or tampering. The system incorporates role-based access control for administrators, encrypted storage for sensitive responses, and secure report generation mechanisms to ensure that only authorized personnel can access aggregated results. To protect against data manipulation and replay attacks, the architecture integrates input validation, session management, and audit logging. Experimental deployment within a university setting demonstrates that the proposed system maintains usability while ensuring confidentiality, authenticity, and integrity of feedback data, thereby enabling trustworthy, data-driven decision-making in higher education.

📚💻 Library Management System — Book Tracking, Member Records & Automated Fine Calculation

This Library Management System is designed to efficiently manage books, member records, and circulation activities within a library. It maintains detailed information on book availability, issued and returned items, and registered library members. The system can automatically track due dates, calculate fines up to a predefined threshold, and generate comprehensive reports for administrative use. By automating core library operations, the system improves accuracy, reduces manual workload, and ensures a streamlined experience for both librarians and members.