Research and Projects

My research interests revolve around Artificial Intelligence, Computer Vision, Statistical Machine Learning, Computational Photography and Computer Graphics.

Some of the projects I have worked on are:

• Exposure Fusion and HDR Imaging: Proposed a flexible Deep Learning based approach for exposure fusion and HDR Imaging from arbitrary number of exposure bracketed shots. Some of the work done in this direction has led to a publication in IEEE International Conference for Computational Photography, 2019.

• Capsule Networks for Optical Flow: Tried to adapt Capsule Nets for optical flow, achieving close to state of the art results on toy datasets.

• Corruption Detection in rendered images using Deep Learning: Created a dataset mimicking common corruptions seen in rendered images like aliasing, pixilation, tearing , etc. and used a VGG-16 based Siamese net to detect them.

• Artistic Style Transfer: Experimented with various losses and architectures for fast style transfer for both images and videos.

• Sound Classification: Used an ensemble of 2D-CNN over log-mel spectrograms and 1D-CNN over raw wave.

• Toxic Comment Classification: Trained an ensemble of NB-SVM and LSTM based methods to detect toxic comments. The solution achieved the top 8% rank in Kaggle competition on the same topic.

• Abstractive Text Summarization: Clustered similar sentences together on the basis of Cosine similarity and K-Means to get word graphs and extracted shortest paths from them, ranking them on the basis of linguistic quality

Apart from AI & ML, I have also dabbled in Compilers, Network Programming, Quantum Computing, and Programming Language Design.