https://elibrary.khec.edu.np/handle/123456789/417 Fri, 17 Jul 2026 04:25:56 GMT 2026-07-17T04:25:56Z https://elibrary.khec.edu.np/handle/123456789/1010 Title: RealTime Emotion Based Music Recommendation System Authors: Asmita Shrestha Jenisha Shrestha Kritima Shrestha Saragam Adhikari (770307) (770315) (770318) (770337) Abstract: The �Real-Time Emotion-Based Music Recommendation System� is an intelligent application that detects human emotions through facial expressions and recommends music accordingly. It integrates computer vision, affective computing, and deep learning to deliver a personalized multimedia experience. The system utilizes a Convolutional Neural Network (CNN) model developed with TensorFlow Keras, using the trained model �emotion-model-7class.h5�, which classifies facial expressions into seven emotions: happy, sad, angry, disgust, fear, neutral, and surprise. Live webcam input is captured and pre-processed (grayscale conversion, resiz ing, normalization), and facial regions are detected using OpenCV�s Haar Cascade Classifier. The CNN performs feature extraction from facial landmarks and classi f ies the emotion. Once detected, the corresponding playlist is selected and played using the python-vlc library, ensuring real-time responsiveness as emotions change. Designed for robust real-world use, the system handles varying lighting, facial orientations, and diverse user profiles. It emphasizes performance, reliability, and cross-platform portability, enabling integration with smart assistants and enter tainment systems. The model was trained using image augmentation, the Adam optimizer, and categorical crossentropy loss over 100 epochs, achieving 60.5% training and 59.9% validation accuracy. The project demonstrates a practical and adaptive approach to real-time emotion-aware computing for personalized music recommendation. Wed, 01 Jan 2025 00:00:00 GMT https://elibrary.khec.edu.np/handle/123456789/1010 2025-01-01T00:00:00Z https://elibrary.khec.edu.np/handle/123456789/1018 Title: FOOD LENS: Food Ingredient Detection & Recipe Generation from Images Authors: Jenish Prajapati Roji Prajapati Shreeya Shrestha Sumina Awa (770314) (770333) (770342) (770345) Abstract: In computer vision, identifying and classifying food from photos is a challenging task. Building on our earlier research on Food recognition, we present �Food Ingredient Detection and Recipe Generation from Images� a sophisticated sys tem that can identify dishes from photos, identify their ingredients, and produce recipes that go with them. Through the integration of food classification, in gredient detection, and recipe generation, this project presents a revolutionary method for cuisine recognition. Our system is made to examine photos of various food, identify the ingredients, and then give users comprehensive recipe instruc tions depending on the components found. The system provides an interactive experience that helps food scholars, and chefs learn about cuisine by utilizing deep learning algorithms and a carefully maintained recipe library. By using cutting-edge AI-driven technologies, this breakthrough helps to preserve and ad vance traditional culinary knowledge. Wed, 01 Jan 2025 00:00:00 GMT https://elibrary.khec.edu.np/handle/123456789/1018 2025-01-01T00:00:00Z https://elibrary.khec.edu.np/handle/123456789/1015 Title: AI Driven Text To Image (Flowers) Authors: Alisha Pokhrel Asmita Ghimire Shiksha Yadav Yogyata Neupane (770301) (770306) (770341) (770347) Abstract: This project explores the use of advanced Generative Adversarial Networks (GANs) for text-to-image(flower) synthesis, aiming to generate high-quality, realistic images from textual descriptions. The project leverages various state-of-the-art GAN models, including StackGAN, AttnGAN, DMGAN, and DFGAN, to compare and evaluate their performance in generating images that accurately reflect the input text. The models utilize techniques such as attention mechanisms and memory networks to improve text-image alignment and image quality. The Adam optimizer is employed for efficient and stable training of the generator and discriminator networks, ensuring fast convergence and high performance. Additionally, an Affine Transformer is used to enhance spatial transformations and preserve consistency between text and image features. The project evaluates the models using quantita tive metrics like Fr�echet Inception Distance (FID), which measures the similarity between real and generated images, and qualitative assessments based on human judgment. The expected outcomes include generating high-resolution, diverse images that align well with the given text, providing valuable insights into the effec tiveness of different GAN architectures for text-to-image synthesis. This research has potential applications in fields such as AI-driven content creation, digital art, and virtual design. Wed, 01 Jan 2025 00:00:00 GMT https://elibrary.khec.edu.np/handle/123456789/1015 2025-01-01T00:00:00Z https://elibrary.khec.edu.np/handle/123456789/1011 Title: GlycoSmart: AI-Driven Nutritional Information Retrieval and Diabetic Health Evaluator Authors: Arbin Thaku Creation Pradhan Prajens Shrestha Yubaraj Poudel (770305) (770309) (770327) (770348) Abstract: Diabetes management demands precise dietary choices, but interpreting nutritional labels poses a major hurdle. GlycoSmart, an AI-powered mobile platform, ad dresses this by automating label detection, extraction, and analysis to offer real-time personalized diabetic health advice. It employs YOLOv8 for efficient object detec tion, attaining an F1 score of 0.97 at a 0.701 confidence threshold and mAP@0.5 of 0.986 for reliable label identification across varied packaging. The Gemini 1.5 Flash OCR model processes extracted data for accurate text recognition in complex formats. The nutritional analysis integrates XGBoost and Random Forest algorithms to predict Glycemic Index (GI) and compute Glycemic Load (GL), with Random Forest outperforming at an R� of 0.9217 and reduced MAE, selected for GI prediction. Built on a Flask backend with Supabase storage and Nutritionix API integration for additional data, the React Native with Expo frontend ensures an intuitive interface. The fine-tuned Diabetica-7B LLM generates context-specific dietary recommendations based on user profiles. By merging object detection, OCR, and predictive modeling, the system provides accurate insights and empowers informed choices for diabetics. The prototype excels in performance and usability but lacks offline mode, AI meal suggestions, and barcode scanning. Future enhancements will boost accessibility and real-world efficacy in diabetes self-management. Wed, 01 Jan 2025 00:00:00 GMT https://elibrary.khec.edu.np/handle/123456789/1011 2025-01-01T00:00:00Z