Overview / Usage

Machine learning (ML) models trained with these datasets (e.g., for medical image analysis) require high accuracy, high confidence levels and high performance to be useful in clinical practice, presenting a significant ML infrastructure challenge for healthcare organizations. In this project, we demonstrate how we leveraged the new Intel Deep Learning Boost (DL Boost) available on the 2nd generation of Intel Scalable processors to optimize the real-time performance of medical image analysis. In particular, we demonstrate 3D-Unet image segmentation on fMRI data from brain imaging.

Methodology / Approach

Low-bit precision interference (INT 8)

Technologies Used

Intel DL Boost (VNNI)