Publications

Automatic mixed precision and distributed data-parallel training for satellite image classification using CNN

Published in Fifteenth International Conference on Machine Vision (ICMV 2022), 2023

This work proposed the implementation of distributed data parallelism (DDP) and automatic mixed precision (AMP) to minimize the memory usage and optimizing usability of GPUs while improving the accuracy of Artificial Neural Network (ANN) model in classifying earth terrains from satellite images

Download here

Seal Capacity Evaluation of CO2 Storage in Carbonate Reef Reservoirs using Equivalent Grain Size Method

Published in 1st EAGE Workshop on Faults in Groundwater, CO2 and Hydrocarbons in Asia Pacific (Canberra, Australia), 2021

This paper discusses our proposed methodology that we (Dr. Kazuo Nakayama in FGI Tokyo and my colleague Kevyn Augusta) developed to evaluate the capacity of seal rock for CCS using the Equivalent Grain Size (EGS) method. Using relation between seal porosity and pore throat size, the EGS method assumes hydrodynamic equilibrium of CO2-gas mixed fluids to calculate each column height in the carbonate reef structure and maximum pressure allowed for CO2 injection.