Naman Jain

I am a final year undergraduate student studying computer science at IIT Bombay. My research interests lie in the field of Machine Learning and Artificial Intelligence, especially in their intersection with Computer Vision.

I am currently working with Prof. Sunita Sarawagi for completing my B.Tech. Thesis. Earlier this summer, I was fortunate enough to get the opportunity to visit University of Maryland as an undergraduate research intern under the guidance of Prof. Abhinava Shrivastava. I also spent an awesome time working with Prof. Arjun Jain in Human Pose Estimation back in 2018.

Email  /  CV  /  Github  /  Google Scholar


  • [Jun 2019]   Visiting University of Maryland as an undergraduate computer vision intern under Prof. Abhinava Shrivastava.
  • [May 2019]   Our paper "On the Robustness of Human Pose Estimation" got accepted in CVPR-W 2019.
  • [Dec 2018]   Recieved Undergraduate research award for work on Adversarial Human Pose Estimation in my Third Year
  • [Nov 2018]   Released code for Pytorch Human Pose Estimation an implementation of various state of the art human pose estimation methods.
  • [Jul 2018]   Released code for my summer project : 3D Pose Estimation from videos using temporal convolutions.
  • [May 2018]   Joined ViGIL Lab, IIT Bombay carrying research in Computer Vision under Prof. Arjun Jain.

segBayes On the Robustness of Human Pose Estimation
Sahil Shah*, Naman Jain*, Abhishek Sharma, Arjun Jain
Workshop On Augmented Human: Human-centric Understanding and 2D/3D Synthesis
Conference on Computer Vision and Pattern Recognition (CVPR 2019)
Arxiv PDF   /   Code

We find the effectiveness of various kinds adversarial attacks in the field on Human Pose Estimation. We perform both image agnostic and universal attacks on five major human pose estimation architectures. We provide an indepth analysis of design choices of pose-estimation systems which would shape future work.

* : Equal Contribution
Selected Projects

For a complete list of projects, please refer to my CV.

19 Pytorch Human Pose Estimation

A PyTorch implementation of various state of the art 2D Human Pose Estimation Techniques along with pretraied models on both MPII & COCO dataset.

obj Temporal Convolutions for Human Pose Estimation in Videos

My research project during Summer'18. We use temporal convolutions (3D Convnets) with initializations from corresponding pretrained 2D CNN for 3D Human Pose Estimation in Videos on Human3.6 dataset. Similar idea for multiperson video prediction - Link

cfd Art Gallery Problem

Implemented approximate solutions for the NP-Hard Art Gallery Problem with O(nlogn) complexity. My first computer science project! (Implemented in Racket)

iitb Undergraduate Teaching Assistant, CS753 Fall 2019

Undergraduate Teaching Assistant, PH107 Fall 2017

Stolen this awesome template from Jon Barron. Thanks alot for sharing!