Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Tracking with Spiking Neural Networks and Event Cameras
This project aims at developing a camera tracking approach with sparse input (events from event cameras) and sparse computation (spiking neural networks). Conventional approaches are built on visual inertial odometry using image and imu data. Ideally, this approach would process image data at high frequency for maximum accuracy. However, this is not attainable on resource constraint devices such as mobile phones or wearables. Event data, in combination with spiking neural networks, can overcome this trade-off by leveraging sparse computation by design. To achieve this goal, we will investigate ego-motion tracking for rotational motion and subsequently investigate 6DoF ego-motion tracking.
This project aims at developing a camera tracking approach with sparse input (events from event cameras) and sparse computation (spiking neural networks). Conventional approaches are built on visual inertial odometry using image and imu data. Ideally, this approach would process image data at high frequency for maximum accuracy. However, this is not attainable on resource constraint devices such as mobile phones or wearables. Event data, in combination with spiking neural networks, can overcome this trade-off by leveraging sparse computation by design. To achieve this goal, we will investigate ego-motion tracking for rotational motion and subsequently investigate 6DoF ego-motion tracking. This project will be done in collaboration with Synsense ( https://www.synsense-neuromorphic.com ) and benefit from their experience as well as our own prior work in this research space.
This project aims at developing a camera tracking approach with sparse input (events from event cameras) and sparse computation (spiking neural networks). Conventional approaches are built on visual inertial odometry using image and imu data. Ideally, this approach would process image data at high frequency for maximum accuracy. However, this is not attainable on resource constraint devices such as mobile phones or wearables. Event data, in combination with spiking neural networks, can overcome this trade-off by leveraging sparse computation by design. To achieve this goal, we will investigate ego-motion tracking for rotational motion and subsequently investigate 6DoF ego-motion tracking. This project will be done in collaboration with Synsense ( https://www.synsense-neuromorphic.com ) and benefit from their experience as well as our own prior work in this research space.