research


Words Matter: Leveraging Individual Text Embeddings for Code Generation in CLIP Test-Time Adaptation | ArXiv 2024

Investigates the role of individual text embeddings in code generation tasks using CLIP for test-time adaptation. Read the preprint: ArXiv.

Do Not Trust What You Trust: Miscalibration in Semi-Supervised Learning | Transactions on Machine Learning Research (TMLR) 2024

Shambhavi Mishra, Balamurali Murugesan, Ismail Ben Ayed, Marco Pedersoli, Jose Dolz
Explores miscalibration in pseudo-label-based SSL methods, exposing overconfident predictions and proposing a novel approach for better-calibrated models. Read the paper: TMLR.

Extended Abstract : TrackRecSign | WiML, NeurIPS 2020

Tracking and Recognition Based Ensemble Approach for Indian Sign Language accepted at Women in Machine Learning Workshop, NeurIPS 2020 held at Vancouver, Canada (Virtually)

Book Chapter : An Insight into Human Pose Estimation and its Applications | Wiley

Shambhavi Mishra CBP Government Engineering College, Delhi,
Janamejaya ChannegowdaRamaiah Institute of Technology, Bangalore,
Kasina Jyothi Swaroop Indian Institute of Technology (Indian School of Mines) Dhanbad

Book Chapter : A Complete Overview on Sign Language Recognition and Translation Systems | Wiley

Kasina Jyothi Swaroop Indian Institute of Technology (Indian School of Mines) Dhanbad,
Janamejaya Channegowda Ramaiah Institute of Technology, Bangalore,
Shambhavi Mishra CBP Government Engineering College, Delhi