I am a PhD candidate in the Department of Computer Science
at The Johns Hopkins University. I am part of the Computational Interaction and Robotics Lab, I am
advised by Dr Bruno Jedynak and Dr Gregory D Hager.
Visual Object Counting: In this work, doubly stochastic Poisson processes and convolutional neural net classifiers are used to estimate the number of instances of an object in an image. We tackle the inherent computational and storage complexity associated with the Cox formulation by employing the Kronecker algebra, taking advantage of the separability of covariance kernels.
Efficient Search: We consider the problem of quickly localizing multiple instances of an object in an image by asking questions of the form ``How many instances are there in this sub-region?", while obtaining noisy answers. We propose a policy for called the dyadic policy for creating image sub-regions and derive analytical expressions for the posterior distribution for this policy, and in turn use this to localize objects.
P. Rajan, Y. Ma, B. Jedynak.
Cox Processes for Counting by Detection.
Journal of Mathematical Imaging and Vision (JMIV), 2018.
P. Rajan, W. Han, R. Sznitman, P. Frazier, B. Jedynak.
Bayesian Multiple Target Localization.
International Conference on Machine Learning (ICML), 2015.
W. Han, P. Rajan, P. Frazier, B. Jedynak.
Bayesian Group Testing Under Sum Observations: A Parallelizable Two-Approximation for Entropy Loss.
IEEE Transactions on Information Theory, 2016.
P. Rajan, P. Burlina, M. Chen, D. Edell, B. Jedynak, N. Mehta, A. Sinha, G. Hager.
Autonomous On-board Near Earth Object Detection.
Applied Imagery and Pattern Recognition (AIPR), 2015.
Y. Otake, S. Leonard, A. Reiter, P. Rajan, J. H. Siewerdsen, G. L. Gallia, M. Ishii, R. H. Taylor, G. D. Hager.
Rendering-based video-CT registration with physical constraints for image-guided endoscopic sinus surgery.
SPIE Medical Imaging, 2015.
P. Rajan, M. Canto, E. Gorospe, A. Almario, A. Kage, Ch. Winter, G. Hager, Th. Wittenberg, and Ch. Munzenmayer.
Automated diagnosis of barrett's esophagus with endoscopic images.
World Congress on Medical Physics & Biomedical Engineering 2009, pages 2189-2192. Springer, September 2009.
S. Seshamani , P. Rajan, R. Kumar, H. Girgis, T. Dassopoulos, G. Mullin, G.D. Hager.
A Meta Registration Framework for Lesion Matching.
Medical Image Computing and Computer-Assisted Intervention (MICCAI). 2009.
R. Kumar, P. Rajan, S. Bejakovic, S. Seshamani, G. Mullin, T. Dassopoulos, G. Hager.
Learning Disease Severity for Capsule Endoscopy Images.
IEEE International Symposium on Biomedical Imaging (ISBI). pp. 1314--1317, 2009.
S. Seshamani, R. Kumar, P. Rajan, S. Bejakovic, G. Mullin, T. Dassopoulos, G.D. Hager.
Detecting Registration Failure.
IEEE International Symposium on Biomedical Imaging (ISBI). pp. 726--729, 2009.
Teaching assistant for the following courses:
Algorithms for Sensor Based Robotics in Spring'18.
Introduction to Computer Vision in Fall'16.
Introduction to Machine Learning in Spring'14.
Asteroid detection: The objective of the project was to develop asteroid detection algorithms that can be hosted on-board a small spacecraft (with limited computational and storage facilities) to serve as a fully autonomous early warning system for asteroids approaching the earth.
Computational Stereo: Stereo matching and correspondence on the da Vinci surgical system.
Barrett's Esophagus: Detection and classification of lesions in the human esophagus using endoscopic images.
Capsule Endoscopy: Capsule Enscopy is used in the diagnosis of Crohn's diesease, intestinal tumors and ulcers.
We developed an ordinal regression framework to rank the lesions in the gastro-intestinal tract using sparse pair-wise preference relationships annotated by the physician.
first name at cs dot jhu dot edu
Hackerman 136, 3400 N Charles Street, Baltimore, MD-21218