Side Projects

BTO Buddy

Description: A one-stop platform for all BTO-related needs with two main components, a chatbot for monitoring of units and a web page for site plan analysis.
Technologies: Python, Chatbot, OpenCV, HTML, CSS, React, Next.js, MongoDB
Blog: Part 1, Part 2
Github Repo: Private

EmoViz

Description: End-to-end classroom analytics solution. It uses a camera in the classroom to captures students’ facial emotions and head pose estimation then converts into actionable insights for teachers
Technologies: Python, OpenCV, PyTorch, HTML, CSS, Javascript
Github Repo: Private

Toaster

Description: A python library that notifies you via channels such as Telegram when your code finishes. This is especially useful for long running codes such as training a neural network.
Technologies: Python
Github Repo: Link

Shirt Similarity

Description: This was created for a talk to introduce participants to computer vision and how to deploy a simple application. Simply take a photo of your shirt and it tells you if you have something similar in your database/wardrobe
Technologies: Python, Keras, OpenCV, Azure Functions
Presentation: Google Slides, Post-talk photos
Github Repo: Link

SureOrNot

Description: Chrome extension which helps to identify whether a particular facebook post is outdated by labeling OLD on it. This is designed to combat fake news.
Technologies: Javascript
Github Repo: Link

MealBot

Description: A telegram bot which prompts family members whether they’re taking meals at home and sends the consolidated results back
Technologies: Python, MongoDB
Blog: Link
Github Repo: Private

Telegram Spam Removal Bot

Description: Telegram bot that helps you to maintain large groups by allowing users to vote if someone should be kicked, giving power back to the community
Technologies: Python
Github Repo: Link

Savings Account Calculator

Description: Calculator to help decide between UOB One, OCBC 365 and DBS Multiplier account
Technologies: Python, HTML, CSS, Javascript
Blog: Link
Github Repo: Link


Hackathons

WSC Hackathon - 2018

Challenge Statement: Automatically generate special effects for sports videos
Solution: We created scripts that were able to detect a player, a basketball and create various effects such as freezing of players, adding names and infographics to videos. Freeze effect was created using Mask-RCNN of players and basketball then offset their mask using optical flow.
Technologies: Python, PyTorch, OpenCV
Demo: Video
Result: 1st Place

Elcothon - 2018

Challenge Statement: Solve a problem in the cinema industry
Solution: We developed an interactive chatbot which encompasses the whole lifecycle of a moviegoer, allowing them to check reviews, purchase tickets, facial recognition walk-in, customising advertisements according to audience age and post-show feedback
Technologies: Python, MongoDB, OpenCV, dlib
Result: 1st Place

Hackommerce - 2018

Challenge Statement: Simplifying Payments and Offline to Online
Solution: We came up with two solutions for each challenge. Firstly for Simplifying Payments, we developed a e-wallet solution which allows users to round up their purchases and deposit small change which can be used for purchases in the ecosystem of merchants. Secondly for Offline to Online, we created a facial recognition tracking platform which assess shoppers facial emotions while browsing store items and convert that into actionable insights. which allows users to narrate a story and AR objects appear accordingly
Technologies: Python, dlib, OpenCV
Result: 1st Place x 2

Samsung Imagination Hackathon - 2018

Challenge Statement: Reimagine the use of the camera
Solution: We came up with Portal, a voice-to-AR app which allows users to narrate a story and AR objects appear accordingly
Technologies: Java, Android, AR
Demo: Video
Result: 3rd Place

DXC Hackathon - The Future of Innovation Challenge 2017

Challenge Statement: Solve any healthcare problem
Solution: We identified the problem whereby upon discharge, doctors do not get any data points of patients with depression till the next followup appointment. We came up with a interactive chatbot for patients to chat with as a companion and it uses the conversational data to automatically fill up PHQ-9 form. This provides doctors with a temporal representation of patient emotional well-being.
Technologies: Cortana, Machine Learning, NLP
Result: 1st Place

BiZiT Big Data Festival 2017

Challenge Statement: Improve customer service using data
Solution: Two-tier dashboard: One for management, one for customer service executive. Management dashboard has overview of all cases and cases resolved by each staff. Customer service executive dashboard has predictive replies for each case based on FAQ knowledge base to speed up response time using Naive Bayes and TF-IDF. Technologies: Python, Machine Learning
Result: 1st Place


If you find any of my projects interesting, feel free to reach out via any medium below!