Spiking Neural Networks for Dynamic Scene Understanding

Dynamic scene understanding will require complex processing of multi-sensor data to achieve both panoptic segmentation and event recognition. Moreover, some applications, such as drone or robotic autonomy, require ultra-low latency, low-power implementations.

The Ph.D. aims to investigate novel approaches, exploring faster, lower computation complexity and efficient sensor signal encoding, challenging the traditional signal processing pipeline. Specifically, we will use spiking neural networks (SNNs) and their intrinsically capability for processing temporal information. Sensor modalities used can include radar, event cameras, and others.

In summary, the candidate will focus on the following research questions:

• Explore novel neural encoding and representation to reduce traditional sensor processing pipelines and extract ideal features for the SNNs
• Investigate SNNs to process information, enabling scene recognition and possibly understanding, requiring multi-object identification and recognition, implying in object segmentation from the sensor data stream.
• Devise a demonstrator to showcase the different components of this research possibly incorporating an end-to-end solution.