About Me

I received my Bachelor of Information Technology (BIT) from the University of Moratuwa in 2017, Bachelor of Laws (LL.B.) from the University of London in 2016, and Bachelor of the Science (Hons) (B.Sc.) from University of Moratuwa in 2018, where I completed my thesis on Capsule Networks advised by Dr. Ranga Rodrigo.

My research interests lie at the intersection of computer vision and deep learning, more specifically in 3D computer vision, image and video synthesis, action recognition and forecasting, video and event understanding, few-shot learning, and self-supervised learning.

Publications

A complete list of publications can be found on my Google Scholar page.
FlexNeRF: Photorealistic Free-viewpoint Rendering of Moving Humans from Sparse Views
Vinoj Jayasundara, Amit Agrawal, Nicolas Heron, Abhinav Shrivastava,
Larry S. Davis
CVPR, 2023
Paper / Project Page

A novel approach which jointly optimizes a canonical time and pose configuration coupled with temporal and cyclic consistency constraints providing high quality outputs as the observed views become sparser.

PatchGame: Learning to Signal in Referential Games
Kamal Gupta, Gowthami Somepalli, Anubhav Gupta, Vinoj Jayasundara, Matthias Zwicker, Abhinav Shrivastava
NeurIPS, 2021
Paper / Project Page / Presentation / Code / Arxiv

Emergent communication via mid-level patches in a referential game played on a large-scale image dataset.

FlowCaps: Optical Flow Estimation with Capsule Networks For Action Recognition
Vinoj Jayasundara, Debaditya Roy, Basura Fernando
WACV, 2021 (Oral)
Paper / Presentation / Arxiv

CapsNet based a finer-grained, motion-specific, better-generalizable optical flow estimation with lesser ground truth data and computational complexity.

PointCaps: Raw point cloud processing using capsule networks with Euclidean distance routing
Dishanika Denipitiyage, Vinoj Jayasundara, Ranga Rodrigo, Chamira U. S. Edussooriya
Journal of Visual Communication and Image Representation (JVCI), 2021
Paper / Arxiv

A novel convolutional capsule architecture with a custom routing algorithm for raw point cloud processing.

Device-free user authentication, activity classification and tracking using passive Wi-Fi sensing: a deep learning-based approach
Vinoj Jayasundara, Hirunima Jayasekara, Tharaka Samarasinghe, Kasun T Hemachandra
IEEE Sensors Journal (Sensors), 2020
Paper

A novel end-to-end deep learning framework that utilizes the changes in orthogonal frequency division multiplexing (OFDM) sub-carrier amplitude information to simultaneously predict the identity, activity and the trajectory of a user and create a user profile that is of similar utility to a one made through a video camera based approach.

DeepCaps: Going Deeper with Capsule Networks
Jathushan Rajasegaran, Vinoj Jayasundara, Sandaru Jayasekara, Hirunima Jayasekara, Suranga Seneviratne, Ranga Rodrigo
CVPR, 2019 (Oral)
Paper / Poster / Presentation / Code

Capsule Networks are cool, but they are shallow. We can increase the depth with the aid of 3D convolutions and skip connections.

TextCaps: Handwritten Character Recognition with Very Small Datasets
Vinoj Jayasundara, Sandaru Jayasekara, Hirunima Jayasekara, Jathushan Rajasegaran, Suranga Seneviratne, Ranga Rodrigo
WACV, 2019 (Oral)
Paper / Poster / Presentation / Code

Capsule Networks can capture the subtle variations present in human handwriting with a very few samples, which can be used to generate new realistic data.

TreeCaps: Tree-Structured Capsule Networks for Program Source Code Processing
Vinoj Jayasundara, Nghi Duy Quoc Bui, Lingxiao Jiang, David Lo
ML for Systems Workshop, NeurIPS Workshops, 2019
Paper / Presentation / Code

Proposing tree-based capsule networks for processing program code in an automated way that encodes code syntactical structures and captures code dependencies more accurately.

TimeCaps: Learning from Time Series Data with Capsule Networks
Hirunima Jayasekara, Vinoj Jayasundara, Jathushan Rajasegaran, Sandaru Jayasekara, Suranga Seneviratne, Ranga Rodrigo
Learning Meaningful Representations of Life Workshop, NeurIPS Workshops, 2019 (Oral)
Paper

Giving CapsNets the ability to understand 1D temporal relationships. We generate capsules along the temporal and channel dimensions creating two temporal feature detectors which learn contrasting relationships.

SmartEmbed: A Tool for Clone and Bug Detection in Smart Contracts through Structural Code Embedding
Zhipeng Gao, Vinoj Jayasundara, Lingxiao Jiang, Xin Xia, David Lo, John Grundy
IEEE International Conference on Software Maintenance and Evolution (ICSME), 2019
Paper / Code

A web service tool, named SMARTEMBED, which can help Solidity developers to find repetitive contract code and clone-related bugs in smart contracts.

Combined Static and Motion Features for Deep-Networks Based Activity Recognition in Videos
Sameera Ramasinghe, Jathushan Rajasegaran, Vinoj Jayasundara, Kanchana Ranasinghe, Ranga Rodrigo Ajith A Pasqual
IEEE Transactions on Circuits and Systems for Video Technology (TCSVT), 2017
Paper

Three schemas for combining static and motion features in videos: based on a variance ratio, principal components, and Cholesky decomposition.

Micro Actions and Deep Static Features for Activity Recognition
Sameera Ramasinghe, Jathushan Rajasegaran, Vinoj Jayasundara, Kanchana Ranasinghe, Ranga Rodrigo Ajith A Pasqual
IEEE Transactions on Circuits and Systems for Video Technology (TCSVT), 2017
Paper / Presentation / Poster

Extracting motion features at micro level, preserving the actor identity, to later obtain a high-level motion descriptor using a probabilistic model.