OpenVINO™ Deep Learning Workbench User Guide

The purpose of this User Guide is to give you instructions on every step of the DL Workbench workflow.

For information on how to start working with DL Workbench, refer to the Get Started documentation.

  1. Obtain Models

    Import original and Open Model Zoo models and convert them to Intermediate Representation format.

  2. Obtain Datasets

    Create and import annotated datasets of different supported formats, upload not annotated dataset, and enlarge datasets using augmentation.

  3. Configure Environment and Work with Remote Targets

    Select a target and device.

  4. Optimize Model Performance

    Accelerate model performance with INT8 Calibration or Winograd Algorithmic Tuning.

  5. Explore Inference Configurations

    Run single or group inference, try different batch and stream combinations to accelerate the performance, visualize model architecture and compare model projects.

  6. Create Accuracy Report

    Measure the accuracy of a model and compare the predictions with the dataset annotations or between optimized and parent models.

  7. Create Deployment Package

    Build your application with Deployment Package and learn how to use batches and streams in your application.

  8. Export Project

    Download an archive with artifacts of your project.

  9. Explore OpenVINO in DL Workbench

    Quick start with the OpenVINO™ toolkit and learn how to use its API and command-line interface (CLI) in the preconfigured environment.

  10. Maintain DL Workbench

    Restore and preserve DL Workbench state and explore options of working with Docker* container.