SeeSay is an affordable, multilingual assistive wearable built on AI smart glasses. A camera, microphone and speakers near the ears let the device sense the world and communicate with the user. In vision mode, the glasses describe the scene on voice command, read printed text such as signs and medicine labels, alert the user to nearby obstacles, and recognise Indian currency notes, with spoken output in Hindi, Marathi and English. In hearing mode, speech is converted to live text captions shown on a small display or a paired phone. The system combines computer vision for detection and text reading, NLP for speech recognition and natural-language output, and time series techniques to smooth detections and use motion data so that alerts are stable and relevant. Lightweight models run on the device, and heavier scene descriptions use a cloud model, with an offline fallback. The goal is a working prototype under ₹10,000 per unit, tested with 10-15 real users through a local NGO or blind school, with the code and dataset released as open source.
The proliferation of deep fake technology, driven by sophisticated deep learning models, poses significant threats in the form of misinformation, identity fraud, and video content manipulation. To address these challenges, this project presents a real-time deep fake detection framework designed to safeguard the integrity of video streaming platforms. The proposed system leverages the combined strengths of Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks effectively record both the spatial capabilities and temporal dynamics of video frames. In particular, RESNET50 was used for robust property extraction, and the LSTM component analyzed sequential patterns to identify anomalies that exhibited deep fake content. This model has been trained and verified using a comprehensive and diverse range of data records, such as FaceForensics++, Deep Fake Detection Challenge (DFDC), and CELEB-DF, to improve your ability to generalize a variety of operating techniques. Performance is evaluated using key metrics including accuracy, accuracy, recall, and F1 scores, with results showing strong identification capabilities. This study highlights the important role of AI in modern video forensics, offering a scalable and efficient solution suitable for integration into platforms like YouTube, Twitch, and Netflix to combat deceptive media in real time.
This project aims to develop an AI-powered platform that helps engineering students manage academics, skills, and time effectively while reducing frustration and burnout. The system will generate personalized study roadmaps and weekly schedules based on the student’s branch, semester, goals (placements, research, GATE, etc.), available study hours, and weak subjects. The platform will recommend resources such as coding sheets, books, and learning materials while also providing productivity and stress analysis through a simple burnout risk scoring mechanism. The goal is to help students balance CGPA, coding, projects, and personal well being in a structured and sustainable way. The project will involve concepts from AI/ML, recommendation systems, backend development, database management, and analytics/dashboard visualization. Expected Features are Personalized academic roadmap, Smart time management planner, Resource recommendation system, Productivity and burnout analysis Progress tracking dashboard. Tech Stack are Python, Scikit-learn, STREAMLIT /React, SQLite/MySQL and optionally backend integration FASTAPI/FLASK Looking for teammates interested in AI/ML, frontend, backend and EdTech related development.
The goal of the project is to build a prototype of a Plant which can desalinate sea-water using solar thermal technology is goes as follows, treated sea water is sent through the Concentrated Solar Plant setup through which the sea-water is converted to steam using concentrated solar light & the sediments stay at the bottom of the tank & the vapours get condensed from steam to get freshwater, the steam moves a turbine to make electricity & finally the sediments are collected to produce salt through this process from sunlight we get electricity, freshwater & salt, the plant runs 24*7 whole year relaying on molten salt method of heat storage or sand battery setup, We are looking on enthusiastic team members willing to work the whole summer with dedication, no prerequisites need pure energy & enthusiasm to build the prototype as the end goal of the project during summer.
This project proposes the development of a machine learning framework to predict human disengagement risk across domains such as employee attrition and student dropout. The core idea is to model behavioral patterns that indicate a decline in participation, enabling early intervention and improved retention strategies. The system will be built using a stacking ensemble architecture composed entirely of tree-based models. At the base level, multiple models such as Decisions Trees, Random Forests and Gradient Boosting methods( XG Boost, Light GBM) will independently learn different aspects of the data, capturing non- linear relationships and feature interactions. A meta learner will then combine these predictions to produce a more robust and accurate final output. The project will focus not only on prediction accuracy but also on interpretability. Feature importance analysis and explainability techniques ( such as SHAP values) will be incorporated to understand the key factors contributing to disengagement. This is particularly important in real-world applications where actionable insights are required by HR teams or educational institutions.
Your Idea Description Appears Here Contact Me in Case of Any queries