There are times when we are left penniless by the end of the month. We used to regret spending so much on things that weren’t worth it. Sounds familiar? Yes, all of us feel that way about something or the other. From electronic gadgets to everyday vegetables, the price varies by a margin. Knowing the price offered everywhere would help us take an informed decision.

Are there instances where you have regretted to have refused a nice offer? For instance, you could have bought a Fossil watch cheaper than the price offered on Black Friday? There could be plenty of reasons, maybe you thought the prices would drop further. Or, somebody would offer at a reduced price ?

What it does

KANJOOS – is a web application that provides the price(s) of an object that you are looking to buy. BAR CODE scanners are there for the same reason right? Yes, BUT we don’t scan the BAR CODES – they’re old school – not the QR codes either (that’s too mainstream :P)

Millions of neurons were at work day and night, training the application. Because of which there are billions of neurons running on Microsoft Azure right now that keeps the application perfect :D Yes, the application is based on Deep-learning !! We believe that this could possibly shape up the e-commerce by paving a way towards a virtual store.

How we built it

On hitting the browser, the app requests you to click a picture of what you want to buy ( Feel free to click a picture of anything around you!! ) As you click and upload a picture of an item, the model that we have trained, learns and classifies the item. It gives us all the details such as the Brand, the Color, and even the Text on the item which our user is interested in buying. Using these details we can scrap the internet and provide them the exact/closest match possible. More importantly, this provides a wonderful opportunity for the companies to understand their users needs and interests better and target them accordingly.

Challenges we ran into

For training the model to achieve 80% accuracy, we had to extensively train the model with varied datasets. The steps involved handpicking the data, pre-processing it. All three of us are Backend brainiacs, but newbies with respect to Android or Front-end development which was quite evident. We even learnt a lot when deploying the application on cloud, as we faced a lot of dependency issues.

Accomplishments that we're proud of

Starting from the daunting task of using Deep learning methods to train a model, and till we successfully deployed it on Azure, it was no easy task! Kudos to the team!! Preparing the datasets to recognize and extract valuable information from images shared by the user, with the clock against us was never easy. Capturing the image using WebcamJS and integrating the model with a Django Application was a great learning experience. Apart from technical skills, we were in awe with the warmth, affection, and the seamless organization skills with which the event took place.

What we learned

It was a great learning experience working on this application development. It gave a real-time work-like experience. We have our android application crash, hence had to build a web-app as a make shift. On the technical front, we learnt to develop Django application, deep-learning techniques on neural networks, web-hosting, and so on. We learnt the difficulties in merging Django and Angular, hosting tensorflow in AWS, and so on as well. The learning has been amazing and the work we completed shows our towards it.

What's next for yo-kanjoos

The project can be expanded to implement Augmented Reality so that one can hover their cam over a product and an overlay appears with real-time price data. The project has wide scope by having both the users and the companies as a potential customer base

Built With

Share this project: