AI applications — including automotive vision and lidar, digital still cameras, surveillance, smartphones — are driving the need for more efficient neural network processing. The trick is getting GPU-level performance within an embedded power and cost budget. This session covers Synopsys’ new Neural Processing Units (NPUs) IP based on a novel architecture and trusted software tools which significantly improve hardware utilization, support the latest neural network trends and scale from battery-powered devices to L3-L5 autonomous driving. We will present inference benchmarks showing performance and power comparisons versus leading GPUs. We will also introduce the evolution of our MetaWare software toolkit to accelerate time to market, and we’ll demonstrate this new NPU solution. 

The neural network architectures used in embedded real-time applications are evolving quickly. Transformers are a leading deep learning approach for natural language processing and other time-dependent, series data applications. Now, transformer-based deep learning network architectures are also being applied to vision applications with state-of-the-art results compared to CNN-based solutions. In this presentation, we will introduce transformers and contrast them with the CNNs commonly used for vision tasks today. We will examine the key features of transformer model architectures and show performance comparisons between transformers and CNNs. We will conclude the presentation with insights on why we think transformers are an important approach for future visual perception tasks.

Time to meet with Synopsys executives and technical experts.

Demo 1: Executing Transformer Neural Networks in ARC NPX6 NPU IP

Demo 2: Driver Management System on ARC EV Processor IP with Visidon

Demo 3: Neural Network-Enhanced Radar Processing on ARC VPX5 DSP with SensorCortek