Authors

Inspiration

Many coronavirus patients develop strong symptoms such as fevers and are unable to call for help. Drone assisted tasks and abilities can be utilized to improve upon rescue missions.

What it does

Drones are used in a combination with a mobile application.  A drone will use machine learning to fly indoors and outdoors while avoiding obstacles. Recorded live streaming will be sent back to the server and saved with the GPS location, time, and date.  The drone can be flown in different areas to scout and scan for unconscious and sick people, detect irregular temperatures using a thermal camera, and report the information. The drone uses face recognition AI technology to extract medical records of the patient quickly. Using the medical and health record, a medical team can know any allergies, medication, health conditions, and other information related to the person. The drone can also drop off medication, first aid, small-sized tools, food and water, and other packages sent by a medical crew if they cannot reach the patient right away.

How we built it

We trained a machine learning model using the TensorFlow library and then used it with the DJI Matrice 100 drone. That model is then stored in the DJI Manifold that is added to the drones. The Manifold is a high-performance embedded computer that can perform complex computing tasks and advanced image processing.

Challenges we ran into

One challenge we faced was lifting and dropping packages safely.

Accomplishments that we are proud of

The drones are able to detect faces with 90-95% accuracy and avoid obstacles such as tree branches and power lines.

What we learned

We learned how to use several technologies such as mapping technologies and many of Amazon Cloud Web Services. We also learned how to control drone hardware using Machine Learning and TensorFlow library.

What's next for CovidScout

We are working on the challenge of drone long-distance travel.

Hackathons

Technologies

amazon-web-services, android, dji-matrice, laravel, python, swift, tensorflow

Devpost Software Identifier

257322