Overview / Usage

With the cost of human labor on the rise and the cost of drones dropping, a number of companies are exploring the use of drones for their last-mile of package deliveries. Current methods being explored include lowering packages via a parachute or string. Alternatively packages are delivered to address with a large backyard. The objective of this project was to develop a proof-of-concept using computer vision technology to detect a safe portion of a driveway at a landed house for the delivery of packages by a drone.

Methodology / Approach

Our final approach was to use convolutional neural networks for semantic segmentation to identify three different classes: driveway, houses, and area of no interest. Once over the target house, our algorithm identifies the driveway closest to the house and iterates towards the driveway for landing.

Technologies Used

Keras, Tensorflow, Robot Operating System, IPython, Python

Other links

• Project Report
• Video