Presented AI technology simplifies synthetic data for GIS digital twins

Satellites Images promise to play a key role in creating digital twins that can help Typical carbon emissionsIncrease agricultural yields and control the oceans. One hope is that the growing volume of satellite data can automatically detect changes, predict trends, and recommend actions.

However, much of this data, in the form of multispectral images, lidar and radar data, outside the scope of a typical human experience, making it difficult for humans to understand and name the data to train better artificial intelligence (AI). Introduction AI It hopes to bridge this gap by combining a synthetic data workflow platform with new GIS (GISIndustry partnerships.

New Partnership with GIS Leader Esri It will make it easy to combine satellite imagery with 3D content to create models for training sets. Another partnership with the Rochester Institute of Technology’s Digital Imaging and Remote Sensing (DIRS) Laboratory will make it easier to bring the labs together. Dersig Satellite synthetic data tools with the cloud-based Rendered.ai platform for bulk synthetic data generation.

“As we move away from images that are easy for humans to interpret, it becomes very difficult to build labeled data sets that can be used to train AI,” said Nathan Konditz, co-founder and CEO of Rendered. “Historically this means that these types of sensors have largely been frozen outside the AI ​​revolution.”

Synthetic data representing different types of satellite data will make it easy to train AI to automatically detect low farm yields, rising carbon dioxide levels, and coral reefs suffering.

The superpowers of synthetic data

The company was founded in 2019 by Nathan Kundtz, Kyu Hwang, Duane Harkness, and Ethan Sharratt. Kondtz said he conceived the idea for the company after being frustrated with the lack of accurate sensor data to train AI while working in the satellite communications industry.

The real sensor data is bias toward common events that are easy to spot or collect, Kondtz said.

“Synthetic data provides an opportunity to overcome biases and gaps in real sensor data and is relatively inexpensive, and encourages innovation and experimentation,” Kondtz said.

The company focused on developing a complete infrastructure for integrating physics-based synthetic data into AI workflows. This includes 3D model sources, simulation tools, content management, managed computation, application deployment, tasks, data source, annotations, and data quality assessment. The platform allows computer vision engineers, synthetic data engineers, and field experts to collaborate and quickly iterate to create data, run scenarios, and strive for improved AI performance.

“Our job was to give these engineers superpowers — and give them the tools to become synthetic data engineers,” Kondtz said.

Overcoming GIS Challenges

Automation is also badly needed by the GIS community, Konditz says, because far more valuable data is being collected than humans can process. Synthetic data workflows for the GIS community can be more challenging than other areas.

“What makes GIS difficult is that the range of things that might need to be simulated is incredibly large – literally the entire planet at many scales and the range of types of sensors used is also very diverse,” Kondtz explained.

For example, a lot of GIS analyzes use sensors that operate outside of what the human eye can see—infrared, radio frequency, and synthetic aperture radar sensing.

Customers generate artificial satellite image data for rare object detection, land cover fragmentation determination and building detectors for asset conditions. Others use the tools to improve the ability to extract data from drone photos, video, and other types of geospatial content.

As this volume of synthetic data continues to grow, automation by companies like Rendered AI offers the potential to apply artificial intelligence to large data sets to solve real-world problems.