I am a Ph.D student at University of Maryland, College Park, working in the Computer Vision Laboratory with Professor Larry Davis. My current research involves using modern computer vision techniques like stacked graph convolutional networks for activity segmentation and deep learning to detect image metadata tampering.
Before this, I completed my M.S in Electrical Engineering at Stanford University , M.Tech at IIT Kanpur and B.E at Bengal Engineering and Science University, Shibpur. I have taken courses based on computer vision, image processing, machine learning, databases and statistics. At Stanford, I was working with Professor Silvio Savarese, on tiny and indistinct object detection and tracking in a crowd.
During the course of my stay in the US for my M.S and Ph.D degrees, I have completed 4 internships at SRI International, One Concern Inc., Meta Co. and Pantry Labs. All of these experiences helped me in developing a strong base in the fields of Machine Learning and Computer Vision.
1. Stacked Spatio-Temporal Graph Convolutional Networks for Action Segmentation
Pallabi Ghosh, Yi Yao, Larry S. Davis, Ajay Divakaran
2. Stacked U-Nets: A No-Frills Approach to Natural Image Segmentation
Sohil Shah, Pallabi Ghosh, Larry S. Davis, Tom Goldstein
3. Understanding Center Loss Based Network for Image Retrieval with Few Training Data
Pallabi Ghosh, Larry S. Davis
ECCV Workshop 2018 from 8-14th September at Munich, Germany
4. Detection of Metadata Tampering through Discrepancy between Image Content and Metadata using Multi-task Deep Learning
Pallabi Ghosh, Bor-Chun Chen, Vlad I. Morariu, Larry S. Davis
CVPR Workshop 2017 from 21-26th July at Honolulu, Hawaii
5. Multistereo System Design
Apoorva Bhatia, Pallabi Ghosh, K.S. Venkatesh
IEEE Second International Conference on Image Information processing (ICIIP 2013) from 9-11th December at Shimla, India
Advisors:Prof. Savarese and Prof. Gordon at Stanford University.
Collaborators: Shabaz Patel and Alexandre Alahi
Develop an algorithm based on neural networks to detect very similar and small objects like ants in extremely crowded environments.
Advisors: Prof. Chichilnisky and Prof. Savarese
Collaboraters: Ayesha Khwaja and John Doherty
We generated an information rich representation of the scene that can be output to the retinal implant that displays scene at a low resolution
Advisor:Prof. Bernd Girod, Dr. Peter Vajda
Collaboraters: Ayesha Khwaja
We took multiple images of a group of people, selected their best faces and replaced these faces in the reference frame to get the perfect picture.
We tried to recognize items placed in a fridge using SIFT matching on segmentation results on difference images.
Advisor: Prof. Brian Wandell
Collaborators: Ayesha Khwaja, Atinuke Ademola-Idowu
We tried to estimate the PSF of camera lenses and use it to deblur captured images.
Advisor: Prof. Andrew Ng
Collaboraters: Nattamon Thavornpitak, Ayesha Khwaja
We tried to estimate stereo map using SVM and K-means algorithm and Daisy Features