ENVI Deep Learning 2.0 Release Notes

Refer to the Deep Learning Help for instructions on using the tools and API. Access the help by selecting Help > Contents from the ENVI menu bar. Then click "ENVI Deep Learning" in the table of contents on the left side of the help page.

 

System Requirements

 

ENVI Deep Learning 2.0 uses TensorFlow version 2.9 and CUDA version 11.2.2, both of which are included in the installation. System requirements are as follows:

To determine if your system meets the requirements for ENVI Deep Learning, start the Deep Learning Guide Map in the ENVI Toolbox. From the Deep Learning Guide Map menu bar, select Tools > Test Installation and Configuration.

This release includes the following new features.

New in this release is ENVI® Machine Learning. You can use the features of Machine Learning with just a standard ENVI license - a Deep Learning license is not required.

With ENVI Machine Learning, you can quickly perform several traditional machine learning algorithms using a variety of one-step classifiers on a single raster. Single-raster classifiers are available for anomaly detection, supervised classification, and unsupervised classification. Additionally, advanced workflows that use multiple rasters as input to training are also available using the ENVI Modeler. This workflow entails extracting data from your raster, generating a trained model, and using that model to perform classification on other rasters.

For details on ENVI Machine Learning and how to use it, open Help > Contents > ENVI Machine Learning from the ENVI main menu.

A new feature in the ENVI Toolbox enables you save the parameter values you set in Deep Learning and Machine Learning tools and reuse those settings in other sessions. How to save and restore parameter values is described in the Help for each tool.

• Base software: ENVI 5.6.3 and the ENVI Deep Learning 2.0 module
• Operating systems:
  • Windows 10 and 11 (Intel/AMD 64-bit)
  • Linux (Intel/AMD 64-bit, kernel 3.10.0 or higher, glibc 2.17 or higher)
• Hardware:
  • NVIDIA graphics card with CUDA Compute Capability version 3.5 to 8.6. See the list of CUDA-enabled GPU cards. A minimum of 8 GB of GPU memory is recommended for optimal performance, particularly when training deep learning models.
  • NVIDIA GPU driver version: Windows 461.33 or higher, Linux 460.32.03 or higher.
  • A CPU with the Advanced Vector Extensions (AVX) instruction set. In general, any CPU after 2011 will contain this instruction set.
  • Intel CPUs are recommended, though not required. They have an optimized Intel Machine Learning library that offers performance gains for certain Machine Learning algorithms.
  • LEARN-546: Classification image failed to display in ArcPro after running TensorFlow Mask Classification.