Baked Potato

Part 2: Dapp
How we plotted da face

Inspiration

What it does

Part 1: Uses machine learning to detect how "under the influence" you are. Pupil dilation, Eye redness, and Eye shape - some of the most conclusive indicators.

Part 2: (we realize that ML's only good based on data that is there). Incentivizes people to contribute to researcher's datasets by anonymously uploading videos/images of themselves "under the influence". Even though anonymous, they can still get paid for their contributions (uses ethereum blockchain in the backend).

How I built it changes to determine how high you are.

Computer Vision : Eye Redness - Locating the eye using Haar Cascade and calculating Redness using Image processing techniques. Pupil dilation- biniary classification with the amazing Microsoft azure ;)
Eye shape - first we plotted points around the eyes. The top of eye, right side and left side. Then we calculated the second derivative as to how those points changed. We fed that into an unsupervised model (PCA dimensionality reduction), and then to a supervised model (SVM) to generate a prediction.

Challenges we ran into

SO. MANY. SOOOO MANY. Our souls broke and then reincarnated. Will fill them in now :) 1) How do we get time series data for the eye shape? Ans: use derivative. How do we get how derivative changes over time? Use second derivative 2) Shortage of Data of people under the influence of drugs like marijuana 3) Metamask had updates that messed up the dapp

Accomplishments that we are proud of

Built an end to end sytstem that captures an image of the face of the subject, sends it to the backend, computes the "How much influenced are you?" using Machine Learning and Computer Vision and displays appropriate alerts based on the result.