Designed to ensure deterministic runtime behavior, CoreAVI’s SafeAICore™ provides a flexible framework for computer vision and neural network inferencing, while achieving the most stringent levels of safety certification. CoreAVI offers the platform building blocks (including COTS-D safety critical hardware designs, software drivers, compute libraries and tools) that enable AI system manufacturers to deploy mission critical pre-trained neural networks as well as implement discreet machine learning algorithms (i.e. support vector machines, random forests) on real time operating systems and to conform to formal safety critical standards (RTCA DO-178C, EUROCAE ED 12C, ISO 26262, and other functional safety standards).
CoreAVI’s SafeAICore framework includes CoreAVI’s safety critical Vulkan SC driver, which provides the foundation for safety critical GPU Compute. GPU Compute is the application of a modern GPU’s data-level parallelism through many-core SIMD engines to solve high volume computations. It is the basis for modern data processing, machine learning and computer vision systems.
Our SafeAICore solution offers the following features and benefits:
- VkCoreVX™ SC – CoreAVI’s safety critical implementation of OpenVX™1.3
- Khronos’ OpenVX industry standard API provides a feature set for implementing and deploying neural networks in safety critical environments.
- VkCoreVX SC is the backbone of CoreAVI’s artificial intelligence and computer vision platform (Safe AI) providing APIs capable of deploying and designing convolutional neural networks, supporting vector machines, gaussian filtering, optimal flow, and much more.
- Built on top of CoreAVI’s safety critical Vulkan SC implementation, providing both graphics and compute capabilities within the same safety critical framework.
- ComputeCore™ – CoreAVI’s safety critical implementation of BLAS (Basic Linear Algebra Subprograms) and FFTs (Fast Fourier Transform) and the mathematical foundation of our Safe AI and computer vision platform.
- Provides safety critical accelerated vector and matrix math operations by implementing the industry standard BLAS functions.
- Provides a safety critical implementation of Fast Fourier Transform.
- Implements Deterministic Vulkan Compute
- The software building blocks are based on flexible GPU compute capabilities and written against a safety critical Vulkan compute API. The advantage of a platform that is written against a safety critical API is that runtime-state-management is deterministic, and no undefined behavior is present at the API level.
- Calculates Worst Case Execution Time
- CoreAVI’s patented safety manager predicts worst case execution time of compute tasks and enables processes to execute within the required time envelope.
- Enables a system integrator to predict the worst-case execution time of executing compute processes in the system.
- Ideal for a wide range of AI and compute-centric applications such as augmented vision systems, object detection and tracking via neural network inferencing, signal processing, image processing for degraded visual environments, security monitoring, encryption, ADAS and more.
- Facilitates the transition to safety certifiable Vulkan Compute from commercial OpenCL® or CUDA® and OpenCV.
- Designed from the ground up for real time and safety certification. Contains no open source components and no 3rd party software.
- Available with CertCore™26262 (ISO 26262 Accredited Safety Assessment Certificate ASIL D) and CertCore™178 (DO-178C / ED-12C Avionics) DAL A safety certification package.
* Our VkCoreVX SC product is based on a published Khronos Specification and is expected to pass the Khronos Conformance Process. Current conformance status can be found at www.khronos.org/conformance.
Don’t see what you need? Please contact us at Sales@CoreAVI.com to determine if we can help support your certifiable application needs.