To find out more about how NSCC’s HPC resources can help you, please contact [email protected].
NSCC NewsBytes June 2022
"NSCC's supercomputing resources have made significant contributions to our research and play an important role in our project's progress. The HPC resources help to speed up the process of both synthetic data generation and neural network training. For example, training a convolutional neural network for classifications of tree health conditions on NSCC is accomplished 15 times faster than what we can do on conventional computer systems. The unprecedented speed for obtaining the results have assisted us in allocating more time to the process of developing and fine-tuning the deep neural networks. NSCC's resources have also helped us generate a large amount of synthetic data, yielding the successful classification of tree health conditions with very high accuracy."
Project Team:
Project Team:
Asst Prof Abdulkadir C. Yucel
Assoc Prof Yee Hui Lee
Dr Shawn Lum
PhD Students: Jiwei Qian, Kaixuan Cheng, Qiqi Dai
For Singapore’s thriving, lush greenery, the number of incidents of falling trees or ‘tree failure’ — the structural deterioration or breakage of any part of a tree — has tumbled by almost nine-fold over the last two decades, from 3,100 in 2000 to 339 in 2020. While it is impossible to completely prevent tree-related incidents, researchers are working on new technologies to minimize the number of incidents.
Radar, ultrasound and electrical resistivity technologies are currently being used for non-destructive tree health assessments to detect structural defects such as cracks, decays and cavities inside tree trunks. However, these technologies are both labour and time-intensive therefore limiting their use for regular health monitoring of trees and may not be optimal for screening tree health on a massive scale.
To overcome the current drawbacks of regular tree health assessments, a team of researchers from Nanyang Technological University (NTU) Singapore’s School of Electrical and Electronic Engineering are tapping on NSCC’s supercomputing resources to develop a deep learning-augmented microwave radar system for rapid detection of tree defects and imaging of tree interiors.
Unlike the current technologies that leverage circular scans around tree trunks, the developed radar allows scans to be performed in a straight trajectory centimetres away from the tree trunks. This reduces the time and labour costs involved in conventional measurement techniques. The system then processes the measured signal strengths through advanced signal processing and deep learning techniques to assess the health condition of the tree, i.e. whether it is healthy or has a cavity or decay inside. Once trees with defects are spotted, the developed radar would then image the tree interior via image processing and deep learning techniques and provide information on the severity of defects, which helps the arborists determine the next stage of action, i.e. whether to treat it or cut it down.
"NSCC's supercomputing resources have made significant contributions to our research and play an important role in our project's progress. The HPC resources help to speed up the process of both synthetic data generation and neural network training. For example, training a convolutional neural network for classifications of tree health conditions on NSCC is accomplished 15 times faster than what we can do on conventional computer systems. The unprecedented speed for obtaining the results have assisted us in allocating more time to the process of developing and fine-tuning the deep neural networks. NSCC's resources have also helped us generate a large amount of synthetic data, yielding the successful classification of tree health conditions with very high accuracy."
Project Team:
Project Team:
Asst Prof Abdulkadir C. Yucel
Assoc Prof Yee Hui Lee
Dr Shawn Lum
PhD Students: Jiwei Qian, Kaixuan Cheng, Qiqi Dai
