Dr. Amar Maharjan

Logo

Resume | LinkedIn | GitHub | Twitter | Google Scholar

Affiliation:
University of North Texas

Contact:
tel: (940)-703-3172
email: amar.maharjan@unt.edu

Welcome

My research interests include point set registration, human pose estimation, semi-supervised learning, computer vision, machine learning, artificial intelligence.

I am a Clinical Assistant Professor at the University of North Texas (UNT), Denton. My research is on Non-rigid registration of point set with large and uneven deformation such as (3D) point sets representing complex articulated human poses. To achieve accurate non-rigid registration in such challenging cases, my work uses both local neighborhood constraint and global shape constraint in the registration. To preserve the local neighborhood structure during registration, I use constraints such as Stochastic Neighbor Embedding (SNE) or Locally Linear Embedding (LLE). I use Keypoint (or landmark) correspondences as a global shape constraint that enforces the alignment of corresponding keypoints between the point sets. To accurately identify the keypoints and their correspondences between the point sets, I use region-aware cluster-based feature descriptor or unsupervised deep learning techniques.

Prior to joining the PhD program at the UNT, I worked in the industry for about 9 years. During that period, I was involved in many projects which includes embedded programming, font (web font) technologies, web development. I have both academic and real-world working experiences in many programming languages (and frameworks) such as C/C++, C#, .NET, Python, PyTorch, Matlab, Objective-C, Java, Javascript etc.

Keywords: Non-rigid registration, large deformation, depth cameras (rgb-d), deep learning, unsupervised learning, local structure, global shape


Publications

Registration of Human Point Set using Automatic Key Point Detection and Region-aware Features
Amar Maharjan, Xiaohui Yuan
Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2022, pp. 741-749.
(paper)
KPIden


Non-rigid point set registration: recent trends and challenges
Xiaohui Yuan, Amar Maharjan
Artificial Intelligence Review, Springer, 56, 4859–4891 (2023).
(paper)


Non-rigid registration of point clouds using landmarks and stochastic neighbor embedding
Amar Maharjan, Xiaohui Yuan, Qiang Lu, Yuqi Fan, and Tian Chen
Journal of Electronic Imaging 30(3), 031202 (5 January 2021).
(paper)
Non-rigid


Point Set Registration of Large Deformation Using Auxiliary Landmarks
Amar Maharjan, Xiaohui Yuan,
2nd International Conference on Urban Intelligence and Applications, 86-98, 2020.
(paper)
LargeDeform


A hybrid framework for automatic joint detection of human poses in depth frames
L. Kong, Xiaohui Yuan, and Amar Maharjan
Pattern Recognition, vol. 77, pp. 216–225, 2018.
(paper)
HumanPoses


Selection of an Intermediate Representation for Program Analysis and Optimization
Amar Maharjan
2008.


Projects

SkyFonts


FontSquare


iManageProject


Customer Relationship Management (CRM)b


Reliability And Monitoring Program (RAMP)


Reliability And Monitoring Program (RAMP)