Publication: COVID-19 identification-based implementation on FPGA using CNN
Loading...
Date
2024-07
Authors
Ammar Hafiz bin Roslli
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Following the recovery from the Coronavirus (Covid-19) pandemic that occurred in 2019, the respiratory disease still remains as a lethal virus, despite
improved treatments. Ongoing efforts to mitigate this harm includes research and development on efficient Covid-19 detection and diagnosis. Reliable methods such as chest radiological imaging and computed tomography techniques are still flawed by manual examinations, being prone to slow detection and human errors. This project presents an accelerated and accurate Covid-19 identifier via the implementation of the AlexNet Convolutional Neural Network (CNN) model on the DE1 SoC FPGA platform. The progression of this work comprises of 10 stages which involves CXR image dataset acquisition, dataset segmentation, Caffe framework installaton, LMDB formatting of dataset, dataset image mean generation, CNN model training, model weights extraction, confusion matrix evaluations, OpenCL environment setup, and PipeCNN compilation. Expected results for this project are quick operation times and favourable accuracy.