MAKER PORTFOLIO
FRUITIPY
[2016]
WebValley Program Overview
At the 2016 Intel International Science & Engineering Fair, I won the Fondazione Bruno Kessler WebValley Fellowship Special Award. Since I recieved the first award, this meant an all-expense paid trip to San Lorenzo (travel included) to attend the three week WebValley summer school. The school was composed of 15 international students, from the US, Ireland, India, and Italy. The first week we had lectures on topics such as spectroscopy, deep learning, and python programming. We spent the next two weeks working on our project, Fruitipy, a portable spectrometer that uses machine learning to determine the ripeness of fruit. Here is the link to the program website.
Project Trailer
This is a video I made describing the project to show to the public. It gives a quick overview in an entertaining manner. It has been used by multiple media sources.
[https://youtu.be/RDv6opPUT4c]
Full Project Presentation Video
This is the video of the livestream of our project presentation to the public of San Lorenzo. It describes all aspects of the project. The presentation starts at minute 1:17:40. I worked on the machine learning aspect, which begins at minute 1:35:30.
[https://youtu.be/sxJwrFtoX6k]
Machine Learning
The photo on the right shows the result of the machine learning I worked on at WebValley. Although I worked on other tasks such as text recognition and image processing, I focused mainly on writing the algorithm to determine how ripe the fruit was. It involved a lot of experimentation with different techniques, such as PCA, SVMs, Random Forests, and QDA. In the end we created a best model generator that would take the data and bin it multiple times (with different numbers of bins) and for each apply LDA, Random Forrests, and SVM. It would then calculate the MCC score and at then report the number of bins and technique that produced the highest result. We were able to get good reults for the different tasks we applied it to. To learn more, watch the presentation video (above) at the specified time.