Delivered Talks

GPU4SNN: GPU-based Acceleration for Spiking Neural Network Simulations

GPU4SNN has been published in PROCEEDINGS OF PPAM 2022 – 14th INTERNATIONAL CONFERENCE ON PARALLEL PROCESSING AND APPLIED MATHEMATICS https://ppam.edu.pl/

GitHub Link: https://github.com/Crypto-TII/GPU4SNN

This videos explain basics of Spiking Neural Networks (SNNs) and Implementation on GPU for GPU4SNN published in PROCEEDINGS OF PPAM 2022. The videos and the repository include CUDA implementation of state-of-the-art (AP, N, S) algorithms and two new proposed (AB, SKL) algorithms for SNN simulation based on a network of randomly connected pulse-coupled Izhikevich neurons.

Introduction to Spiking Neural Networks and GPU Implementation

Delivered a talk at ACCS on “AI and HPC Convergence for Medical Image Analysis”

AI and HPC Convergence for Medical Image Analysis

Efficient deep learning is endorsed for the solution of different problems in medical diagnosis such as enhancement, segmentation, and registration. Nowadays, developing distributed systems for deep learning in medical diagnosis requires analysis of huge amount of data for the training. Under such circumstances, efficient usage of distributed HPC systems avoids delay emerging in training procedure and reduces the overall time. The talk will discuss the impact of the application of AI methods for medical image analysis in distributed HPC systems. The talk will illustrate how these two main fields AI and HPC can be converged together for the real time medical image diagnosis using distributed GPU systems.

Real Time Multi Object Tracking for Skiing on NVIDIA RTX 2080Ti

Using FairMOT – On the Fairness of Detection and Re-Identification in Multiple Object Tracking – Reference Link: https://arxiv.org/abs/2004.01888

Using Yolov3, Tensorflow and Deep SORT – Reference Link: https://github.com/emasterclassacademy/Single-Multiple-Custom-Object-Detection-and-Tracking

Awarded 2nd Prize in the thesis pitch competition at the European Innovation Fest – EIF2020

CNN based Real Time Object Detection on GPU

The following demo is Deep Convolutional Neural Network (CNN) based real time detection of objects on NVIDIA GPU “GeForce GTX 1050” using YOLO (https://pjreddie.com/darknet/yolo/).

Delivered a talk on “U-Net and YOLO for Liver Cancer Diagnosis” at the Faculty Development Program (FDP) conducted by COEP Pune and MTU Imphal, India.

Attended Training Event of HiPerNav on Bio-mechanical Modelling at IHU (Institute of Image-Guided Surgery), INRIA, Strasbourg, France (10 – 12 December, 2018):

Click to access HiPerNav_training_event_in_Strasbourg_v2.pdf

Attended Training Event of HiPerNav on Image Quality Assessment at Paris 13, France (6 – 7 December, 2018):

Click to access Microsoft-Word-Paris-HiperNav-Workshop-Program-26Oct18-Final.docx.pdf

• Awareness of ITU image quality concept and IQA standardisation protocols
• Image quality assessment in the context of medical diagnosis
• Deep learning in medical computer-aided diagnosis

Certificate_Paris

Attended and Assisted Training Event of HiPerNav on HPC and GPU Programming at University of Cordoba, Spain (10 – 17 September, 2018)

Click to access CHW-Full-program-R.pdf

https://www.diariocordoba.com/noticias/cordobalocal/uco-aborda-avances-tecnologia-automatica-extirpacion-tumores_1251451.html?fbclid=IwAR23rfOK5iuDO_yG2quyAHyGRi6zueDDJ3lBjJwPOJgW3t5WHh51o3IgEDY

Visiting Researcher at NTNU Gjøvik, Norway (21st – 30th May 2018)

Implemented high performance Alternating Direction Method of Multipliers for convex optimization problem in CUDA