Vitae

MS by Research in Computer Science and Engineering

International Institute of Information Technology, Hyderabad, India

• Advised by Prof. Makarand Tapaswi.
• GPA: 9.5/10
• Courses taken: Statistical Methods in AI, Data Analytics, Digital Image Processing, Computer Vision, NLP, Multivariate Statistics

BS-MS Dual Degree in Mathematics and Statistics

Indian Instiute of Science Education and Research, Kolkata, India

• MS Advisor: Prof. B. Uma Shankar and Co-advised by Prof. Satyaki Mazumder.
• GPA: 8.38/10
• Courses taken: Regression Analysis, DSA, Statistical Inference, Linear Algebra I and II, Real Analysis I and II, Group and Ring Theory

Graduate Research Assistant - CVIT Lab @ IIIT Hyderabad

International Institute of Information Technology, Hyderabad, India

• MS Advisor: Prof. Makarand Tapaswi.
• PROJECT 1: Long-from video summarization via human-emotion recognition and tracking as well as auto key-points detection in long-form videos like web-series, and movies, which in itself is a novel research problem. It involves a novel way of auto-labeling multi-modal data samples. Further, the approach to fuse multi-modal signals and train our model to make sense of it is itself a very challenging task. Paper arriving soon!
• PROJECT 2: [CVPR'23] - Learning emotion and mental states for movie characters via joint-modeling on scene, characters and dialog modalities to predict more than one emotion or mental state (multi-label) for each character and the whole scene. Besides, we analysed the self-attention mechanism on how model learns to look at relevant modalities at the right time.

MS Thesis @ ISI Kolkata

Indian Statistical Institute, Kolkata, India

• Advisor: Prof. B. Uma Shankar
• PROJECT: Modeling and Reviewing multi-label classification of highly imbalanced Amazon rainforest images via transfer learning using all possible fine-tuned pre-trained models used in ImageNet challenges till date. Plus, implemented & reviewed promising ML models (Random Forest, XGboost, GBC, etc.) on the same.
• Final touch: Ensembled all the above ImageNet models in Integrated Stacking format and fine-tuned (Meta learning) to achieve the best result among all the tried out methods. Article Link

Teaching Assistant @ IISER Kolkata

Indian Institute of Science Education and Research, Kolkata, India

• Head TA for course CS3102 - Programming in Python for PhD and Masters students.
• Head TA for course CS1101 - Introduction to Programming for 2nd year BS students.

How you feelin'? Learning Emotions and Mental States in Movie Scenes

• Authors: Dhruv Srivastava, Aditya Kumar Singh, Prof. Makarand Tapaswi
• Conference: CVPR (June) 2023, Vancouver, Canada.

Multi-Label Classification on Remote-Sensing Images

• Authors: Aditya Kumar Singh, B. Uma Shankar
• Conference: NA. Available on ArXiv.

Video Object Segmentation

Hypernym Discovery

Vehicle Detection and Counter

Detecting Duplicate Quora Question Pairs

• Finalist at QIF (Qualcomm Innovation Fellowship) Indian Version, 2022.

• Winner of ICVGIP 2021 (International Conference on Visual Graphics and Image Processing) Challenge - Audio-Visual Retrieval.

• Awarded the INSPIRE fellowship (2016 - 2021) which is given to attract young talents of the country to pursue and innovate in the field of Science and Technology.