At RSNA 2019, the annual meeting of the Radiological Society of North America, NVIDIA announced updates to the Clara Application Framework that takes healthcare AI to the edge.
The Clara Application Framework includes SDKs to build, adapt and deploy AI powered workflows on NVIDIA EGX, its edge AI computing platform. The latest addition to the framework is the Clara AGX SDK, a domain optimized developer kit for smart devices. You can apply for early access to Clara AGX SDK here. Below is an overview of the updates.
Clara Train SDK
- Privacy preserving Federative Learning, giving data scientist the capability to collaborate and develop increasingly powerful AI algorithms, while keeping patient data private.
- High performing features including Automatic Mixed Precision (AMP) and smart caching techniques provide up to 55x faster performance for training workflows.
- Integrated deterministic training capabilities to guarantee reproducibility for iterative experimentation for data scientists.
- Python based Clara Train API is now available for advanced users to create custom functions and extend the framework.
- The AI-assisted annotation provides new APIs to add 3D polygon editing and continuous learning support. This now makes it possible for medical viewers to not just bring accelerated annotation, but also trigger a training cycle on newly annotated data.
Clara Deploy SDK
- New Clara Pipeline Orchestrator allows developers to get higher performance in a multi-container; multi-pipeline environment
- gRPC based platform APIs that allow the flexibility to extend the platform with newer utilities and allows developers to bring their own services to the platform
- Logging and performance tools provides efficient monitoring of pipelines.
- Efficient memory handling between containers in the same pipeline to accelerate I/O
Clara AGX SDK
- Domain optimized tools to get started on NVIDIA’s Jetson compute platform
- Includes end-end samples for endoscopy and ultrasound video based modalities