PicturePerfect

Inspiration

Data Mining and Information Retrieval are two hot topics in today's tech world and occasionally, pictures might lose some of their essential qualities, such as lusture or hue. So let's say, by chance, a portion of your photo didn't come out as great or probably even came out to be damaged, PicturePerfect can help you retrieve your image in the most probable way.

What it does

The program involves a complex Convolutional Neural Network which breaks down the given photo into a more vast data source, think of 8 as 2 circles, one on top of the other. Then, it notices any unexpected break points, and compares the remaining coordinates with the existing data-set and sees which image matches the test case and then returns the alternate image from the training-set

How I built it

Using Tensor Flow on the Anaconda Terminal(completely written in Python), I trained the program in node.js by providing certain data sets, such as images of triangle, circle, and a rectangle. I used HTML and CSS3 to create the interactive front-end.

Challenges I ran into

1. Setting up Anaconda, learning Machine Learning Algorithms, Create a CSS and HTML5 front-end web-app and trying to figure out how to implement all the countless libraries for the very first time all in just 36 hours!
2. Doing a hackathon alone for the first time.

Accomplishments that I'm proud of

1. Figuring out the basics behind Convolutional Neural Networks and how this theory could potentially be used at a large scale by applying it during imaging processing of more complex test cases such as a scenery.
2. Pulling 2 all-nighters(excluding a much required 2 hour nap)
3. Finishing the project.

What I learned

1. How to multi-task(Point to be noted)
2. How to work on things which was unfamiliar previously
3. How to keep up the motivation even after seeing all prepped-up teams working on amazing projects.

What's next for PicturePerfect

1. Let's keep it a surprise:)

Built With

• css3
• html5
• javascript
• node.js
• python
• tensorflow