Female and Male Scientists Working on their Computers In Big Modern Laboratory. Various Shelves with Beakers, Chemicals and Different Technical Equipment are Visible

AI Connect for Scientific Data

The AiCSD open source sample:

  • Streamlines deployment process and provides the ability to analyze data at the source, improving response times, reducing bandwidth costs, and ensuring data privacy in regulated environments.

  • Employs OpenVINO™ and OVMS for optimizing model performance at the edge, reducing latency, and enabling real-time analysis.

  • Helps research scientists in the fields of chemistry, physics, biology, or environmental science, to expedite discoveries and enhance the precision of their research.

author-image

作者

Overview

AI Connect for Scientific Data (AiCSD) is an open source software sample that connects data from scientific instruments to applicable AI pipelines and runs workloads at the network edge. These pipelines are containerized for efficient deployment, incorporating all necessary components into a standardized, portable unit suitable for edge devices. The reference software contains EdgeX Foundry* based microservices to automatically detect, manage, and process images. The software sample also provides a user interface to manage tasks and jobs.

With this software sample:

  • Software developers can utilize the AI capabilities to add real-time analytics, predictive capabilities, or automated decision-making to their edge device or IoT applications.
  • Data scientists, machine learning engineers, and IoT developers can connect and transfer their machine learning models and pipelines from their training environment to the edge.
  • System integrators can effortlessly integrate AI pipelines into edge computing solutions.

To learn about the high-level architecture and data flow of the AiCSD reference implementation, refer to the Overview.

Features and Benefits

  • Near real time analysis of data generated at the network edge 
  • Bandwidth efficiency and data privacy from processing data locally 
  • Enhanced precision of data analytics to improve the quality of research decisions and outcomes 
  • AI-assisted cell analytics 
  • Seamless transition to integrate models from training to inference at the edge 
  • Broad compatibility with Linux* and Windows*, EdgeX Foundry, OpenVINO™, and Intel® Geti™ AI Platform. AiCSD is based on EdgeX foundry, and leverages OpenVINO for the AI models and is compatible with Intel® Geti™ models.

The Ingredients

Intel provides open source software code that you can readily verify and use when building complete solutions.

The open source sample package includes:

  • Source code
    • Microservices
    • Pretrained models  
  • Documentation
  • Hardware recommendation
  • Tools and libraries
  • Support for Intel architecture-based platforms

Get Started

To start the workflow, choose one of these working models and follow the steps described in the Get Started section on GitHub*.

Two-System Model

  • This model has a Gateway (Edge) system used for processing pipelines and an OEM system used for collecting an input file from a scientic instrument (for example, microscope, scope, and so on). The Gateway system runs Linux, and the OEM system runs either Linux or Windows Subsystem for Linux* (WSL 2). The two systems are on the same network and communicate with each other using a secure SSH Tunnel (port forwarding).

Single-System Model

  • In this model, all the services (intended for the two-system model comprising a Gateway system and OEM system) will run on the same system. The system can be running Linux or WSL 2. This model is primarily used for development or test purposes.