Stash

Inspiration

With thousands of products, hundreds of brands and multiple websites, online shopping can be a cumbersome experience especially when you have to scroll through pages and pages of products only to find a few that you love. It is not only inefficient but also boring. Moreover, for customers who know what they want - colours, silhouettes and brands that fit their style, discovering similar products should be more streamlined and easy.

What it does

This is where Stash comes in - an app that consolidates your wishlists from all the websites you usually shop from, in one place. It notifies you of any changes in price or availability so that you don't have to waste time going through emails notifications or multiple wishlists all the time.

Saw a beautiful blouse at a luxury high-end store but don't want to burn a hole in your pocket? Click a picture using Stash and we'll help you discover similar products in the price range that suits you.

We have three main features namely (1) unified wishlist (2) image search that helps discover better/ similar alternatives, and (3) product tracking i.e. we track changes in availability and/or prices and notify you of the same so that you can make better decisions

How we built it

The image search component to allow for discovery of products was built using Cloud Vision API to label images that users may take from their phone or upload from their devices. These labels can be extracted to search for products on Amazon. We scrape the details for the most relevant products and show them as a list from which you can add items to your wishlist or complete a purchase then and there. Over time, this can be extended to more retailers i.e. we can extract product information from websites like Zara, H&M, etc.

Challenges we ran into

Creating a Machine Learning algorithm that is able to detect products with good accuracy. A simple cloud vision API with limited training was able to predict quite well for most of the objects (scope is limited to clothing products as of now). Over time, as we gather more image data, we can train our model better and extend our functionalitiy to other products possibly.

Another challenge is scraping data on the products. How do we figure out which products are similar on the basis of labels extracted? To overcome this challenge, we can leverage the value we provide to the retailers to develop APIs for them to access their product catalogue.

Accomplishments that we're proud of

Firstly, the idea. Secondly, the fact that we managed to create a prototype and a working model that demonstrates the image search component of the app. Lastly, we're proud of the things we learned throughout the process and the realisation that this is something we would like to continue working on even after the hackathon.

What we learned

The mentoring sessions were really insightful and helpful in terms of addressing different hurdles we came across. So we're thankful for those. And further, we learned how to use a lot of platformss and APIs, etc. as this is our first hackathon for all our team members.