Seminars

View all Seminars  |  Download ICal for this event

Persistent Homology based Topological regularisation for LIDAR Point Clouds

Series: Department Seminar

Speaker: Prahsant Kumar, Ph.D. student, School of Information Technology, Indian Institute of Technology, Delhi,

Date/Time: Sep 27 11:30:00

Location: CSA Lecture Hall (Room No. 112, Ground Floor)

Abstract:
Topological Machine Learning and specifically Persistent Homology (PH),
is an emerging field at the intersection of Topological Data Analysis
and Machine Learning. It quantifies the global shape and evolution of
topological features - connected components, holes, voids, etc. in a
dataset. PH has been used for geometric problems - shape matching,
surface reconstruction, and pose-matching. Gabrielsson et al.[1]
introduce a general purpose topology layer to calculate PH of datasets.
In this talk, I will explain an interesting technique deploying PH
priors on a Graph Neural network network for point clouds. Sparse LiDAR
point clouds cause severe loss of detail of static structures which is
detrimental to navigation. Despite high sparsity, the global topology of
LiDAR outlining the static structures remains intact. We utilize this
property to obtain a backbone skeleton of the LiDAR scan in the form of
a single connected component that is a proxy to its global topology. We
utilize the backbone to augment new points along static topology to
overcome sparsity. We develop a Graph Generative network that is
topologically regularized using 0-dimensional (PH) constraints. This
enables introduction of newer static points along a topologically
consistent global static LiDAR backbone and generation of precise static
points.

Speaker Bio:
Prahsant Kumar is currently a fourth year Ph.D. student at the School of Information Technology at the Indian Institute of Technology, Delhi, India under the supervision of Prof. Prem Kumar Kalra. He is receipient of the Qualcomm Fellowship (2024-25). He has been working at the intersection of point clouds and graphs using tools from Topological Data analysis and algebraic topology.

Host Faculty: Prof. Chiranjib Bhattacharyya