Match ‘N’ Style – A Web-Based Image Processing App for Online Shopping

Keywords: Computer Vision, Web programming, Intelligent Image retrieval

As a sophomore, I conceptualized, designed, and implemented this application within 30 hours during a Hackathon, along with my teammate Budhaditya Bhattacharya. The Hackathon was sponsored by an online e-commerce website named “Reebonz”. The CTO of Reebonz awarded us the first prize and offered us an internship opportunity. Technologies used include PHP, JavaScript, HTML5, and CSS3. Modified Median Cut algorithm was implemented and used for color quantization and dominant color extraction.

You can try the demo here:


These days, many people use online stores to purchase clothes, bags, sunglasses, other accessories, home decorations, etc. Often these stores have countless number of items on sale, which makes it difficult for the person to find exactly what he/she is looking for. These websites usually categorize their products according to brand, type of item, sale events and prices. Some even use “broad” color categories, however, human beings can perceive around 10 million colors, and indeed, it is impossible and not feasible to make 10 million categories. Hence, online stores can only categorize their products (manually) into a few very “broad” categories. Lets say you just bought a beautiful blue dress, and want to find a perfectly matching purse online; how will you do that? At best, you can look at all the “blue” purses the store has to offer, but, guess what… the human eye can perceive over 50 shades of blue! Your dress has a particular shade of blue, you will have search through probably hundreds of products to find the matching color and then, narrow down your search to the bag you like best.

So, we present, “Match ‘n’ Style” – a web-based application (currently only for demo purposes), which aims to enhance the user experience by narrowing down their search, in a way, that the results shown match the color of the picture of a dress/shirt/wallpaper (for home décor) uploaded by the user. The users can also use their webcam to take a snap of the shirt they are wearing. To further help the user, the webcam detects the user’s shirt (or upper part of the worn/held clothes) in real-time and allows the use to capture it.

Poster presented at the Life@NTU Open House for Prospective Students :