Parrot AR Drone 2.0
Drone Image Captures
Facial Recognition Module
What it does
ARDrone 2.0 is an architectural framework aimed towards increasing the ability of develops to produce solutions for drones and other aerial devices. Each modular component sends and receives data via Google's Protocol Buffers to enable development in different programming languages. It also provides real time object(more than 10)detection and facial recognition.
How I built it
We used Python to develop the architectural framework, and developed sample modules to demonstrate the ease of integration into the framework. deep learning models were built in Python using Open-CV, and were trained to associate faces with names and recognize different common objects. We used a custom built library to handle drone inputs and outputs, and built pipelines to facilitate for the real time recognition for the sample modules.
Challenges I ran into
In order to connect to the Parrot AR Drone 2.0, we had to connect to it via Wifi. There was no way to attach an additional wifi-shield, and none of us had an Ethernet port or adapter, and thus we were unable to use various machine learning online APIs for our sample modules. In addition, there was a lot of discussion regarding the development of the architecture in order to find a balance between simplicity and scalability.
What's next for ARDrone2.0
Increase the ease of integration for different modules along with improving the AR Drone's choice of commands when receiving conflicting requests. We also want to develop a software portal to enable users to simply drag and drop code in the corresponding locations to operate a drone in their desired manner.