Details
This project was done as my first undergraduate research project under Prof. Medha Atre, Department of Computer Science and Engineering, IIT Kanpur.
Abstract
This project aims to utilize the the emotional information present in any media to perform cross-modal media retrieval efficiently. We present an implementation to extract this emotion from images as 2-dimensional vectors, and propose two methods to bring this emotion vectors from different medias in the same space, one being a statistical approach, and the other being a learning based approach. We also present a hypothesis, which allows us to establish a ground truth for the cross modal mapping. We perform extensive analysis of one of our proposals, and report the results obtained. We have also proposed and implemented a heuristic to retrieve cross-modal results efficiently.
Amrit Singhal
MS in Machine Learning
My research interests include machine learning, reinforcement learning and artifical intelligence.