Year-2025Year-2025http://repository.iiitd.edu.in/xmlui/handle/123456789/18062026-05-05T11:20:50Z2026-05-05T11:20:50ZApplications of NLP in recipe textsNeeluVaikundam, GurupriyaUpadhyay, RitujBagler, Ganesh (Advisor)http://repository.iiitd.edu.in/xmlui/handle/123456789/19622026-04-23T22:00:23Z2025-07-27T00:00:00ZApplications of NLP in recipe texts
Neelu; Vaikundam, Gurupriya; Upadhyay, Rituj; Bagler, Ganesh (Advisor)
This study addresses the challenge of large-scale, multi-label recipe classification us- ing a real-world dataset of over 600,000 recipes collected from heterogeneous sources. The raw data exhibited significant noise, duplication, and label imbalance, motivating a comprehensive, multi-stage cleaning and preprocessing framework. Key steps included in- gredient normalization, instructions standardization, multi-label parsing, deduplication, and semantic category mapping into hierarchical supercategories. For modeling, we im- plemented a modular pipeline combining TF-IDF feature extraction, classical classifiers, XGBoost, and fine-tuned BERT models to capture both statistical and contextual signals. By adopting a per-supercategory strategy, we minimized cross-domain interference and achieved strong performance, with the fine-tuned BERT classifier attaining a weighted F1-score of 0.7996 and high accuracy on dominant labels. This work demonstrates how rigorous data preparation and modular modeling can enable fine-grained, interpretable recipe classification at scale, providing a robust foundation for downstream culinary ap- plications such as personalized meal planning and intelligent search.
2025-07-27T00:00:00ZTransformer-based models for CNS tumor detection and gradingBeriwal, RohanJana, SagnikSethi, Tavpritesh (Advisor)http://repository.iiitd.edu.in/xmlui/handle/123456789/19582026-04-22T22:00:30Z2025-07-18T00:00:00ZTransformer-based models for CNS tumor detection and grading
Beriwal, Rohan; Jana, Sagnik; Sethi, Tavpritesh (Advisor)
The accurate classification of Central Nervous System (CNS) tumors into their respective sub- types and grades is vital for prognosis, therapeutic decision-making, and patient management. Traditional diagnostic methods, primarily reliant on radiological imaging and histopathology, are time-intensive and prone to inter-observer variability. In this work, we propose a multimodal deep learning framework for the automated detection and characterization of CNS tumors us- ing the AIIMS brain tumor dataset. Our approach leverages a modified CLIP (Contrastive Language–Image Pretraining) architecture tailored for medical imaging, combining a Vision Transformer (ViT) as the image encoder with BioBERT as the textual encoder. This enables robust cross-modal learning between medical images and corresponding textual metadata, such as clinical notes, radiology findings, and histopathological labels.The model is trained using con- trastive learning to align image and text embeddings in a shared latent space, facilitating both image-to-text and text-to-image retrieval.
2025-07-18T00:00:00ZComprehensive analysis of pose estimation and vital sign monitoring systemsLakshaySingh, SarthakGarg, SanyamDeb, Sujay (Advisor)http://repository.iiitd.edu.in/xmlui/handle/123456789/19552026-04-21T22:00:25Z2025-07-18T00:00:00ZComprehensive analysis of pose estimation and vital sign monitoring systems
Lakshay; Singh, Sarthak; Garg, Sanyam; Deb, Sujay (Advisor)
This report provides a comprehensive analysis of modern techniques for contactless health mon- itoring, focusing on the integration of real-time pose estimation and physiological vital sign measurement. It begins by evaluating existing systems for personalized fitness and elderly care, identifying their methodological strengths and limitations. Subsequently, advanced computer vision models and machine learning algorithms for enhancing pose analysis in complex activities like Surya Namaskar and squats are explored. The report details robust camera-based methods for estimating blood pressure and respiration rate, building upon remote photoplethysmography (rPPG). The significance of clinical datasets like MIMIC-III for validating these non-invasive technologies is discussed. Finally, a review of key research papers highlights the synergistic po- tential of combining pose analysis with vital sign monitoring to create holistic, real-time health assessment tools.
2025-07-18T00:00:00ZCampify: a semi-anonymous doubt solving platformRijusmithttp://repository.iiitd.edu.in/xmlui/handle/123456789/19472026-04-21T22:00:09Z2025-07-18T00:00:00ZCampify: a semi-anonymous doubt solving platform
Rijusmit
We present Campify (Entropy), a semi-anonymous, AI-augmented platform designed to reduce hesitation in asking academic doubts among university students. Campify blends verified human mentorship with intelligent assistance to create a safe, context-aware space for doubt resolution. Built using React 19, TypeScript, Tailwind CSS, Node.js, and Firebase, the platform emphasizes dynamic anonymity, verified user roles, and lifecycle-based doubt tracking. It also features an AI-powered assistant to summarize responses and suggest related queries. Backed by survey insights from a sample size of 70 participants, Campify aims to bridge the gap between peer learning, institutional support, and personal comfort.
2025-07-18T00:00:00Z