Silo AI delivers a flexible and robust solution for industrial quality control using deep learning and optimized inferencing on 3rd Gen Intel® Xeon® Scalable processors.
Rule-based quality control that uses specialized sensors to monitor specific characteristics of a process restricts general monitoring and flexibility of an automated quality control system.
- Commodity sensors (e.g., cameras) cannot be used with rule-based systems, preventing widespread adoption of online visual quality control.
- When quality issues emerge, operators must manually review sensor data to understand the underlying cause.
- Rule-based systems are labor-intensive, waste raw materials, and lead to financial losses.
Silo AI provides flexible quality monitoring solutions based on visual computing, deep learning, and commodity sensors.
AI-Driven Monitoring Learns Quality Standards
AI-based quality control enables flexible solutions for a variety of industrial use cases.
Silo AI delivers a flexible and robust solution for industrial quality control using deep learning and optimized inferencing on 3rd Gen Intel® Xeon® Scalable processors. Instead of heuristics from specialized sensors, quality rules are learned from a wide range of visual data. The models are trained on image datasets to automatically determine what “pristine” and “defected” products look like without having to resort to creating handcrafted rules.
Real-Time Data Delivers Real-Time Quality
In the production environment, real-time image data is collected with standard industrial devices, such as RGB or hyperspectral cameras, and inferenced using a highly optimized computer vision model. Data-driven machine learning supports continuous decision making, while the manufacturing process is running. Our AI-based quality solution enables the following:
- Versatile use cases from anomaly detection to multi class defect segmentation.
- Rapid detection of product quality issues and localization of problems to specific sections of the manufacturing process.
- Raw model predictions are turned into domain events, such as quality binning, within a fully configurable decision-making module.
- Quality metrics of interest are presented to production engineers, operators, and managers through real-time and analytics dashboards.
The Silo AI solution is deployable on Intel Xeon Scalable processors for cloud or on-premises environments. Inference components use Intel Distribution for OpenVINO™ toolkit for scalability and migration to other Intel silicon, such as Intel Core™ processors.
Niko Vuokko, PhD
Chief Technology Officer
Intel Optimizations Deliver Up To 62.8X Faster AI Inference Performance
Silo AI completed benchmarking on IntelDevCloud, comparing performance of IntelDistribution of OpenVINO 2021.3 to open source TensorFlow running on 3rd Gen Intel Xeon Scalable processors. Benefits of using the Intel Distribution ofOpenVINO toolkit on 3rd Gen Intel XeonScalable processors include:
- Up to 3.63X more throughput and 19.94X lower latency using the FP32 model
- Up to 5.82X more throughput and62.84X lower latency using the INT8 quantized model
- Supports various Intel based hardware backends at minimal code/configuration change
Succeeding in AI requires a commitment to long-term product development. Let’s start today.