Gesture Recognition: Paper, Scissors, Rock, Shoot!

Description

In our implementation of the Gesture Recognition assignment, we decided to create an interactive game where the user plays several games of paper scissors rock against the machine. The user plays by having his hand in front of the camera, using a special movement to start a round, and simply shooting either paper, scissors, or rock at the correct time in order to compare it against the machine's output and then decide who the winner for the round is.

Breaking it apart

In order to achieve this, we wrote a program with the help of OpenCV in order to analyze the camera's input.
We decided to use color segmentation in order to find the hand's location. After that, we depended a lot on the OpenCV function findContours, which helped us get a clear image of the largest area available after our color segmentation function - the hand. By using this function, we were able to get the hand in a well defined shape.
After having good sense of the hand's location, we decided to go with finger counting, we achieved this by using the function Convex Hull.
Logically, we are calculating the convex hull of our largest contour. Convex Hull ends up returning a polygon from which we deducted that the corners would be seen as fingers. From there, the algorithm is fairly simple, from 0-1 fingers means "rock", from 2-3 means "scissors", and any more than that means "paper".
We also worked on detecting the center of our hand (as shown with the blue circle) in order to be able to track the movements of it. The algorithm for palm center detection finds the defects (the distance between the tip tow and the convex point) and using the defects array, it calculates the palm center.
It gets the rough palm center by getting the average of all defect points. Then it chooses the 3 closest points and gets the circle radius and center which is the palm center.

Light Seeking Solar Panel

Problem

For this project we were assigned to create a basic web server that interacted with a physical computing system. The server was written in C language and it conformed to HTTP/1.x for the purposes of client requests, and it processed client HTTP GET requests for web pages hosted on the server machine. We used sockets to implement the communication between a client on one machine and the server on either the same or a remote machine.

For the physical computing part, we decided to use an Arduino board and with it we built a pseudo solar panel which could be controlled either manually or through the server's interface. An advantage to using the Arduino was that we were able to program some automation to it. The 'Solar Panel' was able to detect the sources of light and adjust its own position according to it.

The logic behind the panel's movement was pretty straight forward. As shown by the second picture, using three different photocell sensors distributed across the board we could create an average which indicated from which position was the light source coming from, readjusting the panel's position in order to get the possible values for each sensor.

Google Search Operators App

Problem

In this side project, my main goal was to get familiar with the development environment for mobile apps, in this case iOS apps. I decided to work on an app that essentially facilitated but more importantly that educated users about the multiple Search Operators that are offered by Google's search engine.

The explanation for this pseudo app is that a user is shown some common search operators which they can use real time in order to have a better search result in their query. Once the user has its query complete, the app will open a webView with the results of their search - hoping to get a better results than a search that doesn't has the search operators.

An explanation of each search operator and its use is given in the More Info button. Ultimately this app would've been a tutorial hoping for users to stop using the app once they got familiar with the search operators.

Structured Face Recognition