-
-
Title Slide
-
UI/UX
-
Logic
-
Hardware/Software
Inspiration
We were inspired by the limitations of conventional home security systems, namely, their inability to distinguish between real threats (such as humans) and false alarms (such as your pet). Our project idea is to create an external home monitoring system that can distinguish between animals and humans. When the system detects motion outside the user’s house, it will trigger, and then determine if the motion was caused by a human or an animal through Image AI, a python package that implements a real-time object detection system. Currently, our competitors have a solution that determines whether something is an animal by analyzing the size. Our solution will improve upon this by identifying what triggered the motion alarm, hence large animals such as deer will not trigger our alarm system.
What it does
Our program has the ability to detect a face with a camera and then sends an email to a user with a recording of the face while it was on display
How we built it
We used OpenCV in Python on a RaspberryPI to implement our facial recognition. Our program was trained with a publically available haar cascade file. When the program detects a face it sends an email to the user.
Challenges we ran into
The processing power of the raspberry pi was our main limitation. Considering that it took multiple hours to install the software to run our facial recognition, actually running the recognition was much harder. We did find that our program works much better on a PC because of this.
Accomplishments that we're proud of
When our program sent the first email with a recording attached we were ecstatic because we had struggled with downloading the libraries for so long. In fact, we were not even sure if the libraries had been installed correctly because we were so accustomed to it not functioning after the long download times. Fortunately, it worked and we were able to see our code function.
What we learned
We learned so much about python and the raspberry pi, as well as how basic AI and email communication works. Specifically, we learned about pip and installing python packages, using the pi terminal, how to send emails in python, and much more.
What's next for AI Safe
In addition to the haar cascade files for facial detection, there is also data for other body parts such as upper body, lower body, arms and legs. We can check for all of these simultaneously to have a much more accurate detection algorithm for a person.
Log in or sign up for Devpost to join the conversation.