Recorded talk

NVIDIA AI & Omniverse: Pegatron Digitalizes AI Smart Factory w/ Isaac Sim, Metropolis, and Omniverse

enplatform captions

Summary

Manufacturers must become software-defined to produce high-quality products faster and cheaper. Pegatron builds a cloud-native Omniverse digital twin that unifies disparate 3D and CAD datasets to give planners and suppliers a real-time view, and uses that twin as a training ground and data factory for perception AIs. NVIDIA stack components—Isaac Sim for robot simulation, Omniverse Replicator for synthetic PCB-A defect data, Metropolis for AOI workflows, TAO for training, and DeepStream for inference—form a pipeline that produced AOI at 99.8% accuracy with 4× throughput. Outcomes include faster factory bring-up, fewer change orders, continuous operational optimization, higher line throughput, and lower costs.

Key technical points

Global problem statement: the $45 trillion manufacturing industry (10 million factories) is racing to become software-defined to improve quality, speed, and cost-efficiency.
Omniverse digital twin unifies disparate 3D and CAD datasets to provide a real-time view of complex factory data to planners and suppliers.
The digital twin is cloud-native and enables planners to optimize factory layout virtually before deploying physical changes.
The cloud-native digital twin doubles as a training ground and data factory for perception AI models.
NVIDIA Isaac Sim (built on Omniverse) is used to simulate and optimize fleets of mobile robots for material movement and pick-and-place robotic arms on production lines.
Automated optical inspection (AOI) points are deployed along production lines to reduce cost and increase line throughput.
NVIDIA Metropolis is used to quickly develop and deploy cloud-native, highly accurate AOI workflows across production lines.
Omniverse Replicator generates synthetic datasets for PCB-A defects that are too complex or costly to capture in the real world—examples cited: scratches and missing or misaligned components.
Training pipeline: synthetic data from Replicator is combined with NVIDIA pre-trained models and NVIDIA TAO for training, adaptation, and optimization of perception models.
Runtime pipeline: NVIDIA DeepStream is used for real-time inference of the trained AOI models in production.
Measured outcome for AOI: 99.8% accuracy and a 4× improvement in throughput after combining synthetic data, pre-trained models, TAO training, and DeepStream inference.
Business/operational outcomes claimed: software-defined factories built on NVIDIA AI and Omniverse can super-accelerate factory bring-up, minimize change orders, continuously optimize operations, maximize production line throughput, and reduce costs.

Engineering relevance

The talk describes an end-to-end engineering stack and data pipeline: ingest and unify disparate 3D and CAD datasets into a cloud-native Omniverse digital twin; use the twin for virtual layout optimization and as a data factory; generate synthetic defect datasets with Omniverse Replicator; simulate robotic fleets and manipulation with Isaac Sim; train and adapt perception models using NVIDIA pre-trained models and TAO; and deploy real-time AOI inference with DeepStream and operationalize AOI workflows via Metropolis. Inputs and outputs explicit in the source: 3D/CAD datasets and synthetic PCB-A defect images in → trained models and AOI inference out. Evaluation metrics provided: AOI accuracy (99.8%) and throughput improvement (4×). The source documents an iterative deployment loop: virtual optimization in the twin before physical changes and reuse of the twin for continuous training/data generation. The talk does not detail guardrails, failure modes, or trade-off analyses beyond the stated metrics and operational benefits.

Japan relevance

For Japanese electronics and discrete manufacturers, the Pegatron example demonstrates a concrete path to digitalize factories: unify CAD and 3D assets into a cloud-native digital twin for planners and suppliers, use simulation (Isaac Sim) to validate robot fleets and manipulators, generate synthetic defect data (Replicator) to close data gaps, and build an integrated training-to-inference pipeline (pre-trained models → TAO → DeepStream) to achieve high-accuracy AOI and higher throughput. The reported metrics (99.8% AOI accuracy, 4× throughput) and the stated outcomes (faster factory bring-up, fewer change orders, continuous optimization, lower costs) are directly relevant performance targets for Japanese production lines considering similar AI-driven modernization.

Full transcript

Transcript

platform captions

The $45 trillion global manufacturing industry is comprised of ten million factories operating 24/7. Enterprises are racing to become software-defined to ensure they can produce high-quality products as quickly and cost-efficiently as possible. Let's see how electronics manufacturer Pegatron uses NVIDIA AI and Omniverse to digitalize their factories. In Omniverse, they start by building a digital twin of their factory, unifying disparate 3D and CAD datasets to provide a real-time view of complex factory data to their planners and suppliers. In the cloud-native digital twin, planners can optimize layout virtually before deploying changes to the real factory. The digital twin is also used as a training ground and data factory for Pegatron's perception AIs. They use NVIDIA Isaac Sim, built on Omniverse, to simulate and optimize their fleet of mobile robots which help move materials throughout the facility... ....as well as the pick-and-place robotic arms that assist on production lines. In the fully operational factory, Pegatron deploys automated optical inspection (AOI) points along their production lines which reduces cost and increases line throughput. NVIDIA Metropolis enables Pegatron to quickly develop and deploy cloud-native highly accurate AOI workflow across their production lines. Omniverse Replicator generates synthetic datasets of PCB-A defects which are too complex and costly to capture in the real world - like scratches and missing or misaligned components. Pegatron then combines the synthetic data with NVIDIA pre-trained models, NVIDIA TAO for training, adaptation and optimization, and NVIDIA Deepstream for real-time inference – resulting in AOI performance that is 99.8% accurate with a 4x improvement in throughput. With software-defined factories built on NVIDIA AI and Omniverse, manufacturers can super-accelerate factory bring up and minimize change orders continuously optimize operations maximize production line throughput all while reducing costs.