How Yeong Used Deep Learning to Create a Successful Data Science Capstone Project!

Capstone Project Insights with Data Science Graduate Yeong Nam Tan

Stay Informed With Our Weekly Newsletter

Receive crucial updates on the ever-evolving landscape of technology and innovation.

By clicking 'Sign Up', I acknowledge that my information will be used in accordance with the Institute of Data's Privacy Policy.

At the Institute of Data, all our students create a capstone project to highlight the knowledge and talents that they’ve gained through their time with us. It’s a perfect opportunity for them to get creative with their new skills, apply them in a business situation and create something that they can bring with them to future job interviews and employment opportunities, as proof of their strengths and abilities.

Data Science & AI graduate, Yeong Nam Tan, shares his insights into his own capstone project. Read on to find out how he went.

Tell us about your capstone project.

My capstone project is entitled “Transforming Tomorrow’s Supermarket Today”. It used Deep Learning to classify images of products and produce picked up by shoppers in a supermarket. With that information, we can understand user preferences and push out information like pricing, availability, country of origin, nutritional value and other bespoke recommendations. All these via a virtual concierge application on the shopper’s smart device. The end game is to improve user shopping experience, and at the same time reduce manpower and real estate costs.

To add a local flavour, I called the project “MAID” which stands for “Marketing Auntie Interactive Dashboard”. 

How did you come up with the idea?

I wanted my capstone project to be based on a subject that is outside my domain knowledge in telecommunication where I spent the last 29 years. I saw this as an opportunity to get out of my comfort zone, and give full rein to my imagination. 

During the COVID pandemic, the supermarket became a favourite destination for many. Who could forget the brouhaha at the supermarket with shoppers elbowing one another to get their hands on toilet rolls! It dawned on me that perhaps a project based on supermarket experience would be interesting. After some research, I realised the supermarket business in Singapore is a fascinating, and lucrative business (14.5% year on year growth!). But it is bedevilled by high manpower and rental cost (78% of opex). How can we leverage technology to do more with less, and transform tomorrow’s supermarket today? That was the genesis for this project.

Where and how did you find the data?

To limit the scope of produce to classify and to ensure sufficient number of images for training and testing, I selected several common household items from a Kaggle dataset.

What challenges did you have along the way?

Deep Learning was only covered towards the end of the course so it took a while to digest and ramp up on my understanding before I was able to apply it. Another challenge was processing the images, which require a faster processor in the form of GPU, and selection of suitable deep learning models. Deployment was also something new to me then, so I had to start learning from scratch.

How did you overcome those challenges?

Lots of prayer! It was an intense two weeks of preparation. I felt alone at times, but am thankful to my trainer for his guidance and new tools for learning today. Google and YouTube became my good friends, and I am also thankful to fellow classmates who encouraged me along the way.

How would you do the project differently if it was within the context of an industry setting for a job?

This is a transformational project. In an industry setting, a task force would be setup with dedicated tasks assigned to each member. One member could specialise in taking lots of images of products and produce from an actual supermarket in Singapore and in different lighting and packaging options. 

Another could specialise in incorporating the image classification algorithm into the cloud and into an app that could be downloaded from an app store. And finally, an overall assessment of the actual business benefits in terms of costs to the company and user overall experiences.

What was the end result?

I was pleased to be able to use several Deep Learning algorithms to classify the images in the end and choose the best for deployment. The whole experience has given me much insight into how to leverage technology for real life applications.

See my capstone presentation on ‘Transforming Tomorrow’s Supermarket Today’ using Artificial Neural Networks for image classification. Other videos can be found on my YouTube channel.

You can also connect with me on LinkedIn!

If you’d like to get your start in Data Science & AI, then book a course consultation today!

Share This

Copy Link to Clipboard