Analysis of stickers to match with a company
In the process of finding a good sticker arrangement
While coming home after a hackathon, hackers are often left sleepy, tired, and flooded with stickers. However, sometimes they care about the arrangements of their stickers, but post-hackathon they are swamped with work and unable to find the time to properly arrange their stickers onto their laptop covers. By generating layouts for hackathon stickers, we wanted to turn the cluttered messes that are the laptops of hackathon goers into something more organized, and find a good use of hackathon stickers.
What it does
Our web app lets you take pictures of your laptop and stickers, and it will find an arrangement that covers a lot of space without overlapping stickers. It also lets you scan stickers and get descriptions about the company it's about, along with links to their home and careers websites.
How we built it
For sticker scanning, we used the Google Cloud Vision API to detect both web entities and character sequences present in the image of a sticker to try to find what company it's for. We then search a database we created of many common companies who give out stickers at hackathons in order to filter out irrelevant results from the api.
To arrange the stickers while taking in account their shape, we used OpenCV to trace the edge of the sticker, then dilated and eroded the image to fix the edge of the image. After that, we produced a filled-vector image by flood filling the fixed edges of an inverted copy of the image.
To retrieve information on the companies, we created a Custom Search Engine and then pulled details from Google Search's API.
Challenges we ran into
Google also deprecated their normal Search API so we had to figure out how to use a Custom Search Engine to make a workaround for it.
Accomplishments that we're proud of
We figured out OpenCV.
What we learned
What's next for StickerStorm
Convolutional neural networks and blockchain as well as an additional, flashy static site.