Yasiru Ranasinghe

I am a 3rd year Ph.D. candidate at Vision & Image Understanding Lab, Johns Hopkins University under the supervision of Dr. Vishal M. Patel. My research focuses on computer vision and its application on crowd analysis, object localization, and representation learning.

I worked as a research assistant for AI4COVID project under the supervision of Prof. Janaka Ekanayake funded by IDRC. During my undergraduate at University of Peradeniya, Sri Lanka, I worked on computer vision and image processing applications on spectral imagery and remote sensing under the supervision of Prof. Roshan Godaliyadda, Prof. Vijitha Herath, and Prof. Parakrama Ekanayake.

Email  /  Google Scholar  /  LinkedIn

profile photo

Research

My research interests primarily revolve around computer vision applications, strongly focusing on using generative models for downstream tasks such as crowd counting, crowd localization, and monocular 3D object detection. I am particularly interested in exploring the potential of diffusion models to enhance the accuracy and efficiency of these tasks, especially when applied to monocular images. My overarching goal is to contribute to the advancement of computer vision techniques that can handle complex scenarios involving crowds and vehicles, with an emphasis on noise reduction, localization, and object detection.


News

  • March, 2024: One Paper accepted at IEEE FG'24.
  • February, 2024: Two Papers accepted at CVPR'24.
  • May, 2023: One Paper accepted by JFST.
  • September, 2022: Joined as a PhD student at Vision & Image Understanding Lab, Johns Hopkins University.
  • January, 2022: One Paper is Accepted by IEEE Access.
  • December, 2021: One Paper is Accepted by IEEE JSTARS.


Selected Publications

CrowdDiff: Multi-hypothesis Crowd Density Estimation using Diffusion Models

Authors : Yasiru Ranasinghe, Nithin Gopalakrishnan Nair, Wele Gedara Chaminda Bandara, and Vishal M. Patel

CVPR, 2024  

The paper proposes a novel approach to perform crowd counting with multi-hypothesis aggregation using denoising diffusion probabilistic models. The approach outperforms existing state-of-the-art methods for crowd counting.

MonoDiff: Monocular 3D Object Detection and Pose Estimation with Diffusion Models

Authors : Yasiru Ranasinghe, Deepti Hegde, and Vishal M. Patel

CVPR, 2024  

The paper proposes using diffusion models to perform monocular 3D object detection and pose estimation. MonoDiff does not require additional modalities to generate intermediate representations to produce box parameters.

Crowd Detection via Point Localization with Diffusion Models

Authors : Yasiru Ranasinghe, and Vishal M. Patel

IEEE FG, 2024  

The paper proposes using diffusion models to localize crowd in images and count the number of people in a scene. The proposed method generates the head locations as a generative task without a separate detector for point proposals.

Transmittance Multispectral Imaging for Reheated Coconut Oil Differentiation

Authors : DYL Ranasinghe , HK Weerasooriya, S Herath, MP Bandara Ekanayake, HMVR Herath, GMRI Godaliyadda, and Terrence Madhujith

IEEE Access, 2022  

The article process a novel non invasive method for food quality analysis, specifically in terms of adulteration, using multispectral images and statistical signal processing and image processing. The proposed method yielded statistically significant results and with significant practical implications.

Constrained Nonnegative Matrix Factorization for Blind Hyperspectral Unmixing Incorporating Endmember Independence

Authors : EMMB Ekanayake, HMHK Weerasooriya, DYL Ranasinghe , S Herath, B Rathnayake, GMRI Godaliyadda, MPB Ekanayake, and HMVR Herath

IEEE Journal of Seleted Topics in Applied Earth Observation and Remote Sensing, 2021  

The paper proposes a novel algorithm to extract endmembers and abundances of hyperspectral remote sensing data. We introduced a regularizer for the nonnegative matrix factorization which improves independence of endmember statistics. The algorithm illustrated superior performance interms of endmber extraction from hyperspectral data.

Convolutional Autoencoder for Blind Hyperspectrl Unmixing

Authors : Yasiru Ranasinghe, Sanjaya Herath, Kavinga Weerasooriya, Mevan Ekanayake, Roshan Godaliyadda, Parakrama Ekanayake, and Vijitha Herath

IEEE International Conference on Industrial and Information Systems, 2020  

The paper process a convolutional autoencoder to realize the nonnegative matrix factorization using deep learning architectures. The proposed methods is applied to extract endmembers and abundances of hyperspectral remote sensing data. The proposed method produced state-of-the-art results on abundance estimation and competitive results in terms of endmember extraction.

Hyperspectral Imaging Based Method to Identify Potential Limestone Deposis

Authors : DYL Ranasinghe, HMS Lakmal, HMHK Weerasooriya, EMMB Ekanayake, GMRI Godaliyadda, HMVR Herath, and MPB Ekanayake

IEEE International Conference on Industrial and Information Systems, 2019  

The paper proposed an algorithm to determine the availability of surface limestone using hyperspectral satellite imagery of an area. We incorporate traditional image and signal processing, and statistical data analysis techniques in the algorithm. Further, we generated a self-supervisied representation for the hyperspectral signature of limestone in the absence of a groundtruth to improve classification accuracy and spatial continuity of the probability map.



Template taken from here. Last updated October 2023.