An example of grids detecting clouds in WorldView imagery.

An example showing where aircraft and helicopters might be.

You might be thinking “grid models are neat, but why should I care?” The answer is in their unmatched efficiency. Grid models can process images in a fraction of the time required by traditional segmentation or object detection models. This speed, combined with our existing models, enable a patent-pending deep learning pipeline that amplifies the value of grids.

Using our optimized classification tools, you can:

• Reduce false positives, to simplify manual efforts to correct deep learning results.
• Save time, which improves productivity and reduces cost for cloud-based processing.
• Improve throughput of hardware-limited applications that cannot automatically scale to meet processing demands.
• Quickly use more complex models and architectures that can better detect unique features, such as the new TensorFlow Optimized Object Classification and TensorFlow Optimized Pixel Classification.

One of the reasons that we bumped the version to 3.0 is that there are some breaking changes with our deep learning APIs. To facilitate a smooth transition, we’ve prepared an in-depth migration guide that is included with the product that helps walk you through: 

• Migrating ENVI Modeler workflows
• Migrating IDL® code using the ENVI Deep Learning API

If you have IDL code that needs to be migrated, you can use IDL for VSCode to run the command “IDL: Migrate Code to Deep Learning 3.0 API” and make changes automatically.

The release of ENVI DL 3.0 is not just an update, it helps bring our deep learning capabilities into the future of remote sensing.

There’s a lot of directions this can go and ways it can be used, including integrated deep learning with ENVI Connect. For those that don’t know, ENVI Connect is an enterprise solution to help teams of analysts work with imagery collaboratively and derive meaningful information.

In the image above, features were automatically detected with ENVI Deep learning 3.0 in 30 seconds. Without grids, the deep learning model used to take 12 minutes to process this scene. With grids it goes from “take a break or multitask” to “have a few sips of coffee and you have your “answer", turning downtime into productivity.

If you are interested in learning more about ENVI DL 3.0 or trying it for yourself, simply send us an email.