-
-
Detective Deadbolt
-
Facial Recognition software in action
Inspiration
The idea for our project was sparked by the issue of lack of security regarding doors in many homes and businesses throughout the United States.
What it does
Our convenient, easy to install, and configurable security system uses advanced neural networks to store and recognize facial image data and vocal audio while corroborating real-time data with existing databases to control door security and log familiar and unfamiliar faces to facilitate future investigation.
How we built it
Starting at the beginning of our time, we quickly started to plan out some interesting ideas. The two major ideas were an AI powered deadbolt and a lie detector. With the ambitious proposal of combining both projects into a deadbolt with facial recognition and voice stress analysis, we quickly started to get to work around 10:30. Almost immediately, we ran into some problems. The program for facial recognition worked well on laptops, but the build dependencies would not install on the Raspberry Pi. Putting this problem to the side, we started designing the hardware parts that would make up the deadbolt. A simple rack and pinion connected to a servo motor would slide the deadbolt into its place, securing the doors shut. While those pieces were printing, we attempted to fix the maddening Python issue. The evening and night of the work time were completely dominated by the problem. At around 2:30 am, the solution finally came, and we rejoiced. After a quick break, the 4 hours that would follow were the most productive part of the entire project. In this short amount of time, we managed to get all the hardware on the door, create and test the software, full assemble the final hack and then do more testing on the completed project. By around 7:00 AM, Sunday, the project was complete, after 20 hours and 30 minutes of little sleep, hard work, and dedication.
Challenges we ran into
Surprisingly, one of our biggest roadblocks ended up being installing the correct version of Python onto our Raspberry Pi. We spent over 13 full hours juggling with different versions of Pip and Python, searching through insufficient documentation, running through different error messages, and manually installing several different programs and files to finally arrive at a maddeningly simple solution. Another challenge we ran into was the way all the components would come together on the door. With limited space and some heavy parts, it took us a little while to figure out how to combine all the parts into a final product.
Accomplishments that we're proud of
After many, many hours of debugging, our Python problem was finally solved. We managed to write a Python program that uses machine learning to recognize certain faces to be allowed through a door, functions as an alarm system for unrecognized and denied faces, and can provide surveillance and security to all doors. All the designed pieces of the hardware portion of the deadbolt worked almost perfectly the first time after printing. Without the need to continuously reprint parts, our time could be used on the software portion of the lock. The hardware portions of the deadbolt came together nearly seamlessly. The wiring for the Raspberry Pi and Arduino is optimized and clean, everything fits together extremely well. Through a remote accessible SSH terminal, we are able to turn on and off the deadbolt wirelessly.
What we learned
We learned the basics and much more regarding the topics of installing and building dependencies needed for the project. Throughout the entire time we spent trying to fix the Python issues, we learned a lot about using wheels as prebuilt modules for Python code, using the terminal to install the dependencies required, and coding for Raspberry Pi in Python. On the hardware side, we learned a bit more about 3D printing, especially things like tolerances for moving parts and designing the prints with the stress it would experience in real life. Overall, we learned about the values of planning and sticking to plans, as well as the value of taking a break once in a while, whether it be for sleep or mac-and-cheese.
What's next for Detective Deadbolt
The current design for Detective Deadbolt is a little basic but leaves us lots of room for improvement in the future. With the addition of some new features, Detective Deadbolt could go from a Hackathon project to a fully functioning product. With the inclusion of a mobile app or GUI, you can access and store faces in order to create a full database of allowed people. With the inclusion of voice recognition and lie detection, we can also implement the ideas we initially scrapped because of time constraints. Finally, with some improved hardware, we can create a more compact and portable deadbolt that will provide security to multiple different types of doors.
Log in or sign up for Devpost to join the conversation.