Inspiration
We first got inspired to make this web app when we discovered Google's database with over 50 million pictures of doodles and we realized what we can do with that and the implementation of machine learning.
What it does
The AAND app takes an image of a doodle provided by the user and using neural networks and the Google doodle database our program can plot and identify any given doodle. The user can go onto our website and upload either a digital doodle or a picture of a drawing and our algorithm will automatically identify the doodle.
How we built it
The AAND app is built mainly on Python for the backend along with several python libraries and modules like OpenCV and Numpy. The frontend is built on the Flask framework, using bootstrap to create a user-friendly and easy user interface.
Challenges we ran into
We ran into some problems with our program misidentifying certain doodles when uploaded. However, with some clever thinking, we managed to locate the problem and fix it.
Accomplishments that we're proud of
The biggest accomplishment in this project was when we figured out how our algorithm could trace the doodles and from the digital drawings, we could progress further into the identification.
What we learned
We learned even more about machine learning and image recognition than we before. Additionally, we also gained experience in extracting data from a database.
What's next for ANND
After this Hackathon, we can continue to refine our algorithm making it even more accurate and even making the website even more aesthetically pleasing.
Log in or sign up for Devpost to join the conversation.