Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Efficient Asynchronous Event-based CNN Processing
Design and implement efficient asynchronous event-based networks to achieve low latency inference.
Keywords: Computer Vision, Event Cameras
One of the amazing properties of event cameras is the high temporal resolution of the visual signal, which is in the range of microseconds. As a result, event cameras do not suffer from motion blur and can capture information in highly dynamic scenes, such as shooting a bullet at a gnome (https://www.youtube.com/watch?v=eomALySSGVU). This makes them extremely promising in critical applications such as autonomous driving. However, it remains challenging to efficiently process the sparse event stream to achieve low latency in vision algorithms.
One of the amazing properties of event cameras is the high temporal resolution of the visual signal, which is in the range of microseconds. As a result, event cameras do not suffer from motion blur and can capture information in highly dynamic scenes, such as shooting a bullet at a gnome (https://www.youtube.com/watch?v=eomALySSGVU). This makes them extremely promising in critical applications such as autonomous driving. However, it remains challenging to efficiently process the sparse event stream to achieve low latency in vision algorithms.
In this project, we seek to either port existing event-based networks to achieve low latency inference or extend a current approach to make it more efficient. The student should have strong self-motivation and should be curious about tackling research challenges in a principled way. Good Python programming skills in one deep learning framework are a must. Please contact us for more details.
In this project, we seek to either port existing event-based networks to achieve low latency inference or extend a current approach to make it more efficient. The student should have strong self-motivation and should be curious about tackling research challenges in a principled way. Good Python programming skills in one deep learning framework are a must. Please contact us for more details.