THU 15:30 - 16:00

This is a 2017 presentation

One of the greatest challenges many artificial intelligence projects face (specifically deep learning related projects) is the high costs associated with hardware. When faced with the problem of vehicle autonomy, most companies resort to high cost industrial grade sensors to tackle the problem, greatly increasing the time-to-market for these systems.

Our research aimed to shift the scale from sensors to pure artificial intelligence, which resulted in a system which can drive a boat autonomously through the canals of Amsterdam using only 5 off-the-shelf standard webcams, completely eliminating the need for sensors like LIDAR. This system, using a new technique dubbed “Reinforced Inference”, was realized in less than 6 months with the deep learning agent itself only needing a small amount of training time (a matter of days) to achieve robustness.

In our presentation we’ll talk about how the project came to be, what techniques were employed and the future timeline for this project. The scientist behind the applied theory will also be available for questions at the end of the presentation.

Presentation highlights:
Project startup and goals
Introduction to Deep Learning
System lay-out and reinforced inference
Routing and object avoidance
Demo and future project timeline

No prior knowledge of machine learning / deep learning is required.

Sign up for this session here.

This session is made possible by Xomnia.