Porting SYCL Accelerated Neural Network Frameworks to Edge Devices

Portable hardware acceleration has become increasingly necessary with the rise of the popularity of edge computing, that encourages data to be processed and stored as close to the source of origination as possible. This capability is needed in areas where bandwidth and latency are restricted and network stability, privacy, or security are unreliable or insecure.

A representative use case is unmanned aerial vehicles (UAVs) where size, weight and power (SWaP) constraints are a key constraint. In this work we leverage a NVidia Jetson edge device as an onboard accelerator to improve the perception and situational awareness of UAVs through an end-to-end deep learning framework to perform multi-object semantic segmentation of radar aerial point clouds.

To achieve this, we modified the PointNet architecture incorporating insights from the aerial domain to create a new SYCL backend using the open source DPC++ compiler. We were able to achieve comparable performance with the native CUDA backend used by ONNX. By modifying the architecture we are able to use any accelerator including NVidia for the edge case.

Download Presentation

×


Register for the oneAPI DevSummit hosted by UXL:

Register Now