Partnership with GE Ventures' Avatar Systems

NVIDIA announced a collaboration with GE Ventures' Avitas Systems. Through this collaboration, the two companies will apply cutting-edge AI computing technologies to industrial equipment defect detection and inspection robots that utilize advanced data analytics.

Avitas Systems can target specific points for inspection and develop image and video data collection paths applicable to various robotics such as drones, crawler robots, and autonomous underwater vehicles (AUVs).

Driven by 3D modeling, these paths are repeatable at the same angle and location. Based on this repeatability, a wide range of images captured over time are input into Avitas Systems' cloud-based platform, where advanced image analysis is performed to measure precise defects in industrial equipment, such as cracks and corrosion.

The platform assesses the severity of even the most subtle defects, often difficult to detect with the human eye, helping customers determine when equipment needs to be replaced and address potential issues early.

Avitas Systems is leveraging NVIDIA DGX-1 and DGX Station to train artificial intelligence for automated defect detection. Avitas Systems data scientists built generative adversarial networks (GANs) to minimize the labeling work of captured images and convolutional neural networks (CNNs) for image classification.

Avitas Systems stores deep learning models within its innovative AI Workbench solution, which processes the data in real time and allows the models to be retrained and applied to new use cases.

“Our collaboration with NVIDIA has enabled us to fully commercialize our state-of-the-art, self-service AI Workbench, and we look forward to expanding its capabilities with the new NVIDIA DGX Stations based on Volta,” said Alex Tepper, founder and president of Avitas Systems. “With Workbench, Avitas engineers can easily develop and deploy new deep learning models to train software that automatically recognizes defects in inspection sites.”