Inspiration
As two out of the four members of our group are international exchange students, our group was inspired to combine our desire to learn more about mobile developments and web programming into an app inspired by a popular book and movie series in the UK. However, we wanted to add a little twist by incorporating an unexpected element in the form of a phone prank.
What it does
The first part of our program is aimed at matching users with one of the harry potter characters which they most resemble. This can be done either by uploading an image or using the system’s camera. The user is then able to stop and the image or camera input analysis and receive a call to a number previously provided. This call is a fun prank call which corresponds to the character which matched closes to the image or camera input.
How we built it
This program was built using Selenium, Teaching Machine website API, and the Twilio API. Fir the initial portion of the project, we focused on creating a data set of images for the main characters in Harry Potter. To do this we first used Selenium to create a program for crawling the images from google for each of the characters. This set of images then had to be filtered so that the data was more accurate and cropped to focus on the facial features. Once the data set had been generated, the website, API Teaching Machine, to train a machine learning algorithm and generate a classification model differentiating the characters facial features. This data could then be used on the website to take the user input and find the character with the best match. Our website was designed using HTML and JavaScript. On the back end, Node.js was used to create a local server on which we could run the website.
Challenges we ran into
One of the biggest challenges we ran into was creating a data set which was not massively biased towards an individual character. One of the main factors in this bias was large differences in the colors of image sets, and in order to combat this, each image had it’s background removed and the image was normalized. This was majorly successful, however we began running into issues again as we increased the number of characters in the dataset and a new bias began to emerge.
Accomplishments that we're proud of
We are extremely proud of not only being able to call a user from our website, but also being able to generate a message which corresponds to the results of the classification data.
What we learned
We were able to learn a lot about the Twilio API and gain experience in working with unfamiliar APIS’s.
Log in or sign up for Devpost to join the conversation.