Enview

AI Hypervisor: Scalable ML framework for rapidly deploying custom 3D AIs

AI models work best when they are trained on data that is as similar as possible to your real world data in application. The AI hypervisor approach many factors related to your data and its attributes, then selects the optimized model match for your job. This approach allows top tier performance models to always be applied appropriately.

Performance optimization

  • Models’ selection is based on point cloud density, sensor type, georegistration and more.
  • Combination of machine learning and automatically selected algorithmic techniques that extract maximum value from the data available.
  • Ever-expanding number of models and training sets for novel use cases and situations.

Enview Orchestrator: Programmable engine for automating 3D analysis workflows

Enview’s orchestrator is a cutting edge workflow engine, designed to flexibly meet the needs of any industry or use case. This allows rapid deployment of new, complex, and multi-step workflows to accommodate novel use cases for point cloud geospatial analysis.

Automate many types of workflows

  • Conditional branch logic – modify, improve, transform, or combine geospatial data in an automated way.
  • Human machine workflow optimization – Human in the loop QA and correction.
  • Fully automated computation – execute complex computation and inference.

Take advantage of parallel compute

  • Subdivide large geospatial data into small area tasks and execute in parallel.
  • Take on virtually limitless scale projects.
  • Execute in record time.

Manage via external API

  • Deploy new workers and workflows into a running system with no downtime.
  • Write workers in any language that Docker supports.
  • Run jobs on any EC2 machine.
  • Leverage spot instances to reduce compute costs.

Synthetic Training Data: Massively accelerated creating new 3D AIs and models

Create synthetic point clouds of any CAD model object

  • Convert any 3D CAD model into a consistent point cloud for placement in scenes.
  • Add noise and variation to the point cloud objects.
  • Model rare objects or difficult to observe combinations of objects.

Develop synthetic scenes

  • Develop synthetic scene models of procedurally generated terrains.
  • Build synthetic scenes that add variation, or situation specific elements.
  • Add any number of objects to scenes.
  • Objects can be rotated, tiled, translates, scaled or resampled to match their environment.
  • Object placement can be random or in specific patterns to better simulate real world data.

Train models and workflows faster

  • Build systems on generated data to avoid the time and cost of data collection.
  • Design workflows to address specific edge case scenarios that are difficult to collect.
  • Save labeling time by creating scenes that are annotated with semantic segmentation, instance segmentation, and 3D bounding boxes automatically.

Worker Library: Pre built automation blocks

We have built a comprehensive library of worker operations covering most common operations that various use cases require. This provides reliability and speed in any use case we approach.

  • ML Training – These workers are used to initiate training one of Enview’s Machine Learning Models using data sets appropriate for that Model.
  • ML Execution – These workers execute one of Enview’s Machine Learning Models on a data set.
  • Computation – Workers in this category perform computational tasks, creating new data sets from existing data sets.

  • Visualization – Workers in this category are used to prepare acceleration structures needed to view large data sets on the web.
  • ETL – Workers to extract, translate and load data sets include data preparation, cropping, splitting, filtering and packaging.
  • Custom – If you can’t find what you need then Enview supports the addition of custom workers by our partners.

Enview Explore: Democratizing complex 3D data analytics for end-users

Enview Explore is a web application that uses AI and cloud computing to automate LiDAR classification, segmentation, analysis and extraction. Enview Explore utilizes a novel approach to applying AI to 3D data that yields faster results than traditional LiDAR software.

Classification & Segmentation
Terrain Modeling
Change Detection
Feature Extraction
Foliage Penetration

Classification & Segmentation

Leverage the power of the cloud to execute complex analysis and render classified data in a fraction of the time it takes for traditional manual analysis.

Terrain Modeling

Easily turn massive LiDAR point clouds into highly-accurate and actionable digital surface and digital terrain models. No desktop software necessary.

Change Detection

Quickly highlight meaningful change between two datasets over time. Repeat the process as often as needed to monitor regulatory compliance or understand patterns of life.

Feature Extraction

Leverage AI and machine learning to power automated extraction of unique objects in both 2D and 3D formats.

Foliage Penetration

Enview’s unique 3D AI performs foliage penetration and hard object detection below the canopy in a matter of minutes, not days.