Zoom Room 1

Group A (E5b)

Vision-Based Intelligence towards Self-Driving Vehicles

Chua Yong Liang¹, Sampson Bong Shung Hoong¹, Libut Jaztynn Phoemela Barairo², Marcelo H. Ang Jr.³, Chong Yue Linn³, Li Zhikai³

¹SRP Student, River Valley High School

²SRP Student, Anglo-Chinese Junior College

³Advanced Robotics Centre, Faculty of Engineering, National University of Singapore

Abstract

This project aims to algorithms for a remote-controlled car to achieve self-driving capabilities. Algorithms are based on deep-learning and inspired by how humans learn. The vehicle used is a radio-remote-controlled car using a Raspberry Pi 4, with a camera sensor. Different deep learning architectures, namely the Linear, Recurrent Neural Network (RNN) and Latent deep neural network models were implemented on the radio-remote-controlled car in Python using the Keras library. These models were trained using a dataset of camera images together with throttle and steering values collected during manual drives on a fixed teardrop-shaped track. The models were compared to determine the best-performing model in avoiding obstacles while following a track with the most accurate model being the Latent model.

Group B (E5c)

An Investigation of an Underactuated Propeller for Unmanned Aerial Vehicles

Jonathan Chew Jian Pin¹, Wang Zixun¹ and Ang Marcelo H. Jr.²

¹SRP Students, River Valley High School

² Department of Mechanical Engineering, National University of Singapore

Abstract

A novel single-rotor tailsitter was designed, prototyped and tested. This project focuses on the hovering mode of the tailsitter. The tailsitter uses an underactuated propeller for pitch and roll control, and two control surfaces for yaw control. This design aims to reduce the number of actuators and motors as compared to conventional tailsitters that use two or more brushless motors and servo motors, thus reducing cost and simplifying maintenance. Several experiments were carried out to investigate the effect of various parameters of the underactuated propeller on the rolling moment produced. The tailsitter was then tested on a 2-axis test stand, to visualise the pitching and rolling of the vehicle. A proportional-integral-derivative (PID) controller was used to stabilize the tailsitter. The tailsitter demonstrated momentary hovering.

Zoom Room 2

Group A (S6)

Quotient and Remainder of Polynomials

Panchal Garima Lala¹, Yap Von Bing²

¹SRP Student, Catholic Junior College

² Department of Statistics and Applied Probability, National University of Singapore

Abstract

In this paper, we will show the proof of uniqueness and existence of a quotient and remainder in division of integers and polynomials. We will show that when a polynomial p(x) is divided by a linear polynomial x–a, the remainder is p(a), as stated by the Remainder Theorem, and moreover the quotient is a certain symmetric polynomial in x and a.

Group B (4c)

AI post-impressionism art of protein structures using Neural style transfer deep learning method

PL Sriram¹, Dr Chen Yu Zong²

¹SRP Student, Anglo-Chinese School (Independent)

²Department of Pharmacy, National University of Singapore

Abstract

Artificial Intelligence (AI) Neural Style Transfer is an evolving deep learning method which is widely used in many parts of the world to create new artworks. The Neural style transfer method works by using an optimisation technique to combine the style statistics of an image and content statistics of another image to create a new artwork. In this project, the style used for neural style transfer would be post impression art where the final artwork would have the key features of this art form. Protein structures were also employed in the creation of this post-impressionism artwork. The artwork aims to reflect the devastating impacts of pollution of nature and how pollution affects the human body and the proteins in it. Hence, protein structures were manipulated using software to create the artwork. This project was completed using various AI and visualising software, namely Tensorflow, Adobe Photoshop and Biovia Discovery studio.

Zoom Room 3

Group A (S4a)

AI Impressionism Art of Medicines

Huan Yaohong¹, Chen Yu Zong²

¹SRP Student, Dunman High School

²Department of Pharmacy, National University of Singapore

Abstract

This project aimed to create an impressionist artwork using the chemical structures of medicines, aiming to portray Alzheimer’s disease and its impacts in a visual way to advertise and engage the populace. This was achieved through first creating content of artwork by piecing up the chemical structures of the medicines and substances that made up the body. Then an impressionist painting related to the theme of artwork was chosen and its style was applied to the artwork through Neural Style Transfer. It was found that altering the values of style weight, content weight and total variation weight during the machine learning process can affect the styles of the final artwork.

Group B (A2b)

Optical Technique for Rapid and Quantitative Bacteria Detection

Sheng Jiarui¹, Yakovlev Nikolai Leopoldovich²

¹SRP Student, Dunman High School

²Institute of Materials Research and Engineering, A*STAR

Abstract

Bacteria detection plays an important role in different sectors like clinical diagnosis, food industry and environmental quality control. Portable, rapid and cost-effective methods are still challenging to achieve. This study investigates the performance of a novel optical technique for detecting bacteria in environmental water samples, specifically from cooling towers. The technique involves treating bacteria captured by filter membranes with a proprietor reagent that changes colour in contact with bacteria and measuring the colour change using a colorimeter. Orientations of filter, power sources and environmental conditions were found to be the parameters influencing the measurement of the colorimeter. The procedure of the test was evaluated and improved based on the problems observed. The test results and of this colorimetric technique were compared with the standard plate count results, and a model of the correlation was presented.

Zoom Room 4

Group A (S2b)

Comparison of Microbial Safety and Quality Indicators in Spontaneous v.s. Probiotic-Controlled Fermentation of Carrots in a Home Fermentation Setting

Phua Jia Yang¹, Tan Jia Jun Shaun¹, Ten Mei Zhen Michelle², Li Dan³

¹ SRP student, Hwa Chong Institution (College Section)

² Research Assistant, Department of Food Science & Technology, National University of Singapore

³Assistant Professor, Department of Food Science & Technology, National University of Singapore

Abstract

This project aims to decipher if the addition of Lactobacillus plantarum 299v probiotic strain to control the fermentation of carrots in a home setting can result in a more consistent product with improved microbial safety, and health benefits as opposed to those of spontaneous fermentation, without compromising the organoleptic properties of the carrots. Through obtaining various data such as pH, texture, colour and microflora count through conducting microbial, physical and chemical analysis on the fermented carrots, we were able to determine that probiotic-controlled fermentation allows for enhanced microbial safety as well as faster fermentation, as compared to spontaneous fermentation. Therefore, we can successfully recommend to growing home-fermentation enthusiasts the use of commercially available starter culture probiotic strain L. plantarum 299v as a superior and novel alternative to spontaneous vegetable fermentation.

Group B (S1b)

Do species traits influence the outcomes of coral interspecific interactions?

Goh Tian Hui (Rachael)¹, Huang Danwei²

¹SRP student, Hwa Chong Institution

²Department of Biological Sciences, National University of Singapore

Abstract

Due to space constraints on the tropical reef ecosystem, corals compete between species to gain optimal amounts of sunlight and nutrients for survival. Competitive interactions include overgrowth and direct interaction, when corals use allelopathic chemicals or sweeper tentacles to gain ground over other species. In this study, we investigated whether differences in evolutionary history and life history traits affect outcomes of interspecific competition. We modelled these effects and individual trait differences against competitive outcomes using linear regression and t-tests. Results show that phylogenetic distance is a significant factor in influencing the outcomes of coral-coral competition, but trait differences as a whole do not affect competitive outcomes. However, among specific coral traits, growth rate, coloniality, symbiont transfer and sexual system are significant factors in competition. Possible explanations are given to explain this significance, but more research is needed to confirm our findings due to the complex relationship between traits and competition.

Zoom Room 5

Group A (M4)

Limb Hypothermia for Preventing Chemotherapy Induced Peripheral Neuropathy: Effects of Different Limb Hypothermia Modalities on Motor Nerve Functions

Angelica Lee¹, Tia Shi Ting¹, Raghav Sundar^2,4, Darren Tan², Chin-Tan Wei Lin³, Aishwarya Bandla²

¹SRP student, Eunoia Junior College

²The N.1 Institute for Health, National University of Singapore

³SRP teacher advisor, Eunoia Junior College

⁴Department of Haematology-Oncology, National University Hospital, Singapore

(Abstract withheld from publication)

Group B (E4)

Designing a Low-cost and Compact Circuit to Effectively

Perform Nuclear Quadrupole Resonance (NQR) Spectroscopy on Medical Pills Containing N-14 Nuclei

Tay Kar Ray¹, Yong Ethan¹, Heng Chun Huat²

¹SRP Student, Raffles Institution

²Department of Electrical and Computer Engineering, National University of Singapore,

Abstract

This paper aims to present the design of a low-cost, low-power circuit that can perform Nuclear Quadrupole Resonance (NQR) spectroscopy on paracetamol pills by activation of the nitrogen-14 nuclei present within the pills. As traditional NQR spectroscopy is known to be costly and requires high power, this research will aim to create something portable that can be done within the constraints of a simple laboratory. In this project, a successful transmitter circuit has been designed to perform NQR spectroscopy using the spin-lock-spin-echo (SLSE) pulse sequence. However, due to time constraints, the receiver involving low noise amplifier is not included.

Zoom Room 6

Group A (E5g)

Computational Modelling of Soft Robots

Seah Yue Min Vanessa¹, Xu Yi², Chirikjian Scott Gregory³

¹SRP student, Nanyang Junior College

²2nd year graduate student, National University of Singapore

³Department of Mechanical Engineering, National University of Singapore

Abstract

This project aims to explore designs of soft robots as well as application of Matlab to plot and observe the curvatures of soft robots. Matlab is a high-performance language which assists with the coding of soft robots.

Soft robots are made of highly compliant materials like rubber, allowing them to be much more flexible. Soft robots are useful because they are flexible enough to deal with very fragile materials. This can be seen from soft robotic grippers being used to pick up fragile items like uncooked chicken eggs without damaging it.

Group B (M6a)

Transcriptional Candidates as Biomarkers for Allergic Rhinitis

Agrawal Praakhar¹, Lee, Anika W Xuen¹, Venkatesan Subasriram¹, Dr Anand Kumar Andiappan²,

Asst Prof Elizabeth Huiwen Tham³, Dr Abigayle Ng⁴

¹SRP Student, Raffles Institution

²Singapore Immunology Network, Agency for Science Technology and Research

³Department of Paediatrics, Yong Loo Lin School of Medicine, National University of Singapore

⁴Teacher-Mentor (Biology and Research), Raffles Institution

Abstract

This study focuses on identifying transcriptional candidates as biomarkers to elucidate Allergic Rhinitis (AR) among the paediatric population in Singapore, through their differential expression and correlations with other genes. Our study population revealed 8 statistically significant and immunologically relevant gene biomarkers, known to be present in the adult cohort, namely: IL5RA, ALOX15, PTGDR2, P2RY2, PRSS33, TFF3, OLIG2 and HRASLS5. From which, TFF3, OLIG2, P2RY2 and PRSS33 were shortlisted among the 8 genes for further research. We explored their possible mode of action in relation to AR pathophysiology and tested them against the Asthmatic (As) population to ensure their specificity to AR only. Overall, we concluded that OLIG2 is a possible biomarker for AR in the Singaporean paediatric cohort.

Zoom Room 7

Group A (S3c)

The Square Root Function: History and Computation

Tan Tee Khoon Nicholas

SRP student, Victoria Junior College

Abstract

This project aims to investigate the origin of various iterative methods used by mathematicians to calculate the value of square roots, studying the history of these methods, which include the Babylonian method, the Bakhshali method (Indian method), the digit-by-digit calculation (Chinese method), as well as the Newton-Raphson method, and determining the most effective method for calculations through numerical experiments. It was found that iterations using the Bakhshali method yielded results that converged the fastest to the true square root value, due to its quartically convergent algorithm, meaning that the number of correct digits of the approximation roughly quadruples with each iteration. On the other hand, the digit-by-digit calculation had the slowest convergence rate, with an increase of only one correct digit with each successive iteration. The present study also delves explicitly into the irrationality of the square root of 2, including two proofs of irrationality, one analytic and the other geometric.

Group B (S1a)

The Evolution of Colours in Butterflies

Ess Alexander Jones Shijie¹, Lee Yuan Xi², Zoey Lau², Ian Z.W. Chan³

¹SRP Student, Temasek Junior College

²SRP Student, Raffles Institution

³Department of Biological Sciences, National University of Singapore

Abstract

Butterflies display sexual dichromatism, differences in the colouration of male and female wing patterns. These colours act as communication signals, aiding in attracting mates, camouflaging or intra-species recognition. However, much remains unknown about butterfly wing chromaticity, such as the mechanism by which colours appear, and the nature of the relationship between sex and wing colouration. We analysed images of the dorsal and ventral wing surfaces of the males and females of 268 species of butterflies from the Nymphalidae family to extract colour data. An ancestral character reconstruction was then conducted to investigate the relationship between colour and sex, and examine when different colours--blue, purple and red--appeared in the Nymphalidae phylogeny. We demonstrate that in species with fewer colours, females exhibit more colours than males, while the opposite is true in species that exhibit more colours. In addition, it was found that colours appear via both evolutionary radiation and independent evolution.

Zoom Room 8

Group A (B1b)

Liar Liar Pants on Fire: A Computer Vision Approach to Deception Detection

Nathaniel Tan Xin Rui¹, Lim Qin Zhi², Rui Jue²,

¹SRP Student, Raffles Institution

²DSO National Laboratories

Abstract

Lie detection has been a subject of interest due to the severe repercussions it can have on society especially in high-stakes situations. With the advancement of computer vision and machine learning algorithms, we present a visual approach to lie detection where facial characteristics are used as features to train our model. Our model is a stacking ensemble, consisting of Random Forest, XGBoost, and a Neural Network as base classifiers and a Random Forest as the meta classifier. Our model managed to achieve an accuracy of 86.96% and AUC of 0.844 on a state of the art court dataset and an accuracy of 80.00% and AUC of 0.761 on a novel politician dataset.

Group B (B1c)

Improved Graphical User Interface for Interactive Pixel Tight Segmentation

Jotham Lim¹, Nathan Tew¹, Xin Yan Goh², Ching Lik Teo²

¹SRP Student, Raffles Institution

²DSO National Laboratories

Abstract

Recent advancements in interactive segmentation mainly focus on algorithmic aspects of the process. However, there is little work in development of an effective graphical user interface (GUI), which this work aims to produce. We propose 4 modifications to a basic GUI (by Sofiiuk et al., 2020): 1) an integrated labelling system, 2) the use of a confidence heatmap, 3) the use of bounding boxes and 4) the use of scribbles. We evaluated modifications 1-3 by creating custom auto clickers to emulate the clicking strategies a human would employ using the modifications. Compared to the standard clicking strategy, our results show that the use of a confidence heatmap yields marginal improvements for unseen datasets, while the use of bounding boxes and scribbles showed general improvements. However, the effects of retraining the prediction models on appropriate clicking strategies, as well as the actual performance with actual human annotators, remains to be investigated.

Zoom Room 9

Group A (S3b)

Joseph Fourier, the Fourier Series, and Fourier Analysis

Chua, Richelle Aubrey Ang¹, Chua Seng Kee²

¹SRP Student, St. Andrew’s Junior College

²Faculty of Science, National University of Singapore

Abstract

This paper provides a brief introduction to the Fourier Series, Fourier Analysis, and some of Joseph Fourier’s contributions. Here, we discuss some of the basic properties of the Fourier Series such as its pointwise convergence. We also tackle some real world applications of Fourier Analysis in the fields of Physics and data management. This is meant to target high school students who are interested in engineering or mathematics at a higher level.

Group B (S2a)

Developing Vegan Fish Balls from Soy-based Protein Isolate and Wheat Gluten

Nicole Ang Rong Chen¹, Lee Siu Ching², Ran Xinli³, Assoc Prof Yang Hongshun⁴

^1,2SRP student, Raffles Junior College

³PhD candidate, Department of Food Science and Technology, National University of Singapore

⁴Associate Professor, Department of Food Science and Technology, National University of Singapore

Abstract

Increasing environmental concerns about overfishing and health concerns regarding toxins in contaminated fish has led to an increased demand for vegan options for alt-fish products. Contributing to this demand is also the popularity of fish products and in particular, fish balls in Asian countries. This paper thus studies the use of soy protein isolate (SPI) and wheat gluten balls as a viable vegetarian alternative to real fish balls through investigation of its rheological and physicochemical properties.

Zoom Room 10

Group A (M2b)

Patient Perspectives on the Utilization of Telehealth in Cardiac Rehabilitation during COVID Pandemic

Desai Krishi¹, Anbarasan Denishkrshna², Lim Serene Peiying³, Kayambu Geetha³, Gani Qamaruzaman Syed³, Poojari Shefali³, Lee Si Hooi³, Wee Candice³, Lai Shuet Ming⁴, Ng Hsuen Nin⁴, Ong Mui Cheu⁴, Ooi Lean See⁴, Chong Wooi Fuon⁴, Koh Karen⁴, Yeo Tee Joo⁴ (main mentor and supervisor)

¹SRP student, Yishun Innova Junior College

²Faculty of Science undergraduate, National University of Singapore

³Departmart of Rehabilitation, National University Hospital Singapore

⁴Department of Cardiology, National University Heart Centre Singapore

Abstract

Our study aimed to assess patients’ impressions of a telehealth-based hybrid cardiac rehabilitation (CR) programme during the COVID-19 pandemic. Each hybrid CR programme comprised one orientation videoconferencing session, two fitness assessments (baseline and upon completion), five in-person exercise classes and two teleconsultation sessions. Patients assessed the programme upon completion via online questionnaires. 18 patients underwent the hybrid CR programme. The majority were male (n = 14, 77.8%), aged between 51 to 65 years (n = 10, 55.6%) and used a wide range of mobile applications (n = 14, 77.8%). 100% of patients agreed/strongly agreed that the programme was easy to use, as valuable as in-person sessions, saved time and had lower infection risk compared to in-person sessions. Overall ratings for videoconferencing, teleconsultation and the hybrid programme were 9, 8.6 and 9.3 (out of 10) respectively. Telehealth-based CR is feasible, acceptable and should be considered as an alternative CR delivery method.

Group B (A5)

Human Movement Analysis using Handphone IMU sensors

Amy Ling Yuhui¹, Natasha Yeo Ting Yu¹, Tan Shannon¹, Chin Zheng Yang², Ang Kai Keng²

¹SRP student, Raffles Institution

²Institute for Infocomm Research A*STAR

Abstract

This study explores the use of data from embedded accelerometer and gyroscope sensors in smartphones for Human Activity Recognition (HAR). Data processing and classification was conducted on 3 pre-existing public datasets. Since the intrinsic differences of each dataset led to varied classification accuracy results, an investigation was carried out to postulate this relationship. To supplement this, self-collection of data with 3 subjects was carried out using AndroidHAR, a self-developed Android application with features specific to data collection needs. AndroidHAR’s design facilitates data processing and increases the possible future applications of AndroidHAR for other studies that require HAR data collection. This study utilises Machine Learning to classify activities. Multiple classifiers and various hyperparameters were tested on validation sets, before being applied on test sets to obtain HAR accuracy. Results are generally acceptable but more robust methods to process data and classify activities can be explored.