Research
Publications
2021
Beyond Classification: Knowledge Distillation using Multi-Object Impressions
British Machine Vision Conference (BMVC)
View Paper2020
ManufacturingNet: A Machine Learning Toolbox for Engineers
ML4Eng Workshop at Neural Information Processing Systems (NeurIPS)
View PaperAcademic Experience
May 2020 — Aug 2020
Graduate Research Assistant @ Carnegie Mellon University
Pittsburgh, PA
- Developed a method to classify abnormal ciliary motion, using two-stream convolutional neural networks. Obtained a marginal improvement on the current benchmark performance for binary classification.
- Contributed to ManufacturingNet, an open-source Python library offering robust implementations of featurization and learning algorithms to solve data-driven problems in the manufacturing industry. Accepted to NeurIPS 2020’s workshop on Machine Learning for Engineering Modelling, Simulation, and Design.
Jul 2019 — Dec 2019
Project Assistant @ Indian Institute of Science
Bengaluru, India
- Implemented a novel approach for data-free deep learning using knowledge distillation for object detection.
- Built a baseline model to generate an object detection dataset from a trained model’s weights for the Pascal dataset. The dataset was then used to distill knowledge from a larger model to a smaller model.