News
- Feb 2024: Summer internship opportunity: please visit this page
- Feb 2023: Paper accepted at the CVPR 2023 Conference (CORE A* ranked).
|
People
- Saumyaranjan Mohanty - Ph.D. (External from DRDO)
- Madhumitha V - Ph.D.
- Harsh Udai - Ph.D.
- Naveen George - M.Tech (3Y)
- Sunayna Padhye - M.Tech (3Y)
- Roshin Roy - M. Tech (2Y)
- Rupa Kumari - M. Tech (2Y)
- Anish Pawar - M. Tech (2Y)
- Puneet Rajan - M. Tech (2Y)
Alumni can be found here.
|
Research
Following is an approximate clustering and labeling of the research (click on the label to find relevant research).
|
|
Long-Tailed Training Data
Real-world datasets suffer skewed label frequency distribution, generally with a long-tail. Models trained on such data generalize poorly. We aim to
contribute effective solutions to alleviate the adverse effects casued by class imbalance in the training data.
|
|
Data Engineering for Deep Learning
While the heaps of digital data surely serve the data hungry deep learning, it comes with a set of new challenges (e.g., data redundancy, complexity of training, etc.). We aim to investigate engineering solutions to these data and learning related challenges.
|
|
Robustness
DNNs are vulnerable to adversarial samples that are a dangerous threat for deploying these models in practice. Therefore, the effect of adversarial perturbations warrants
the need for in depth analysis of this subject.
|
|
Adaptability
Deep Learning has been data and resource intensive. However, real-world may challenge us with various constraints to apply these sophisticated tools.
Adapting deep learning techniques/models (e.g. knowledge transfer, domain adaptation) across tasks and to challenging environments such as low data
and data-free scenarios is of high importance.
|
|
Interpretability
NNss are complex ML systems. Because of end-to-end nature of their learning, these models suffer from lesser
decomposability and hence many of us treat them as black-boxes. We study these models in order to make their inference more human-interpretable
and explainable, and devise useful inferences and tools.
|
|
ML/AI in diverse disciplines
ML and AI are generic set of tools that can be applied to solve problems from diverse set of fields.
This is a list of projects that I had a chance to apply the ML/AI techniques in areas other than CV, NLP, and Speech. [Figure taken from teachingai.blog]
|
|