Projects
Data Augmentation using Part-based Deformations of Shapes (Matlab, Torch)
Done as part of my thesis research work. This project addresses the problem of learning good image classifiers with limited labeled data. We propose to use the part analysis of shapes extracted from available images to augment the labeled data. We show improved performance with CNN classifiers for shape classification tasks using such augmentation.
View Publication
Personality Analysis from Inerview Videos (Lua, Torch)
Done as part of "First Impressions" challenge organized by Chalearn Looking at People, where we designed and implemented a Deep Learning based solution for Personality Analysis from Videos. We employed a novel training technique for videos and scored 2nd rank in the competition.
View Project View Publication
Video and Image inpainting (Python, Pytorch)
Done as part of challenges organized by Chalearn Looking at People on Image and Video Inpainting, where we employed GAN based techniques to remove captions from videos and images. We scored 3rd rank in the competition.
View Project 1 2 View Publication 1 2