Stitch

Stitch

A photo filter app that uses neural networks
Summary:

The application allows for users to upload two images from their phones, or from the web. These images are sent to a server, where a convolutional neural network was used to attempt to recreate the first photo in the style of the second photo. The resulting photo is displayed to the user, and is saved in a gallery. Running the neural network required an immense amount of graphics processing power. For this reason, AWS servers with powerful hardware were used. Also utilized torch and loadcaffe to configure the parameters of the neural network. Afterwards, created a simple API with node.js, which allowed for photos to be uploaded from other devices, as well as to send generated photos back to the devices. The final piece was an iPhone application that acted as the user interface.

Hackathon:

LA Hacks 2016

Awards:

Third Place

Platform:

iOS, Web

Built with:

Swift, Lua, node.js, neural networks, machine learning, torch, loadcaffe, aws

Website

Check it out on Devpost

Source Code:

Check it out on Github

Sample Creations:

_config.yml Screenshot 1

_config.yml Screenshot 2

_config.yml Screenshot 3

_config.yml Screenshot 4

_config.yml Screenshot 5

_config.yml Screenshot 6