This is a project I worked on during November 26th ~ 17th, 2019. At an AI object tracking hackathon.
- Detecting a human target (that is nearest to the drone camera's center screen) using 'MobileNet' neural network.
- Track that person while orbiting 'counter-clockwise', with the help of PID controller.
Note! This project uses 'openVINO' inference engine / library. Which is basically a framework developed by Intel, that can optimize models trained in other frameworks (ex. Caffe, Tensorflow) for use on Intel CPU / GPU. Model parameters are already in this Repository, but you'll prbably need to get used to using openVINO if it's your first time.
'w/s/a/d' to control the XY motion
'key up/down/left/right' to control throttle and yaw, respectively.
Python modules : tellopy, numpy, av, cv2, simple_pid, openvino.inference_engine
python3 telloCV.py
This will bring up a pop-up screen, and from there, the program will try to connect to the drone itself, make sure WiFi is connected. When the video can be seen consistently, you're safe to start flying!
The code is heavily based on a telloCV project.