Machine Learning
What Is Machine Learning?
Machine Learning (ML) involves algorithms that analyze data, identify patterns, and make decisions with minimal human intervention. Unlike rule-based systems, ML models adapt through exposure to new information, improving accuracy over time. NYRIS employs deep learning—a specialized ML technique using neural networks—to process visual data from industrial equipment, retail shelves, and e-commerce platforms.
Analyze Your Use Case
NYRIS trains ML models on synthetic datasets generated from CAD files, enabling precise identification of machinery parts and retail products. This approach reduces reliance on real-world data collection, cutting deployment times by 70% for partners like Daimler and METRO.
How Machine Learning Works
NYRIS’s ML framework operates through a three-phase workflow optimized for industrial applications:
- Data Preparation and Synthetic Generation: The system ingests CAD models, technical manuals, and product images, then generates photorealistic synthetic data using GANs (Generative Adversarial Networks). For automotive partner Renault, this process created 5 million training images from 500 CAD files, simulating varied lighting and wear conditions.
- Model Training and Validation: Convolutional Neural Networks (CNNs) analyze visual features like edges, textures, and shapes. NYRIS’s proprietary algorithms achieve 99.3% accuracy in classifying industrial components, validated against real-world datasets from DMG Mori’s production lines.
- Deployment and Continuous Learning: Trained models integrate with SAP ERP systems via APIs, enabling real-time updates to inventory databases. Post-deployment, federated learning techniques allow models to improve across client networks without sharing sensitive data—critical for IKEA’s global retail operations.
Industrial Applications
Manufacturing Automation
NYRIS’s ML models reduce machine downtime by 72% for DMG Mori through predictive maintenance. Vibration and thermal data from CNC machines are analyzed to forecast bearing failures 48 hours in advance, enabling proactive replacements.
E-commerce Personalization
IKEA’s visual search platform, powered by NYRIS, uses ML to recommend products based on style and spatial compatibility. Customer conversion rates increased by 35% as the system learned from 2.8 million user interactions monthly.
Retail Inventory Management
METRO’s shelf-scanning robots leverage ML to track 50,000 SKUs with 98% accuracy. The system detects stockouts and misplaced items, automatically updating inventory records and reducing manual audits by 85%.
Benefits For Your Company
Eliminate Manual Inspections:
Automate quality control with ML models that detect defects at 60 frames per second, achieving Six Sigma defect rates below 3.4 per million.
Accelerate Time-to-Market:
Train models on synthetic CAD data to deploy solutions 10x faster than industry benchmarks, as demonstrated by Trumpf’s laser cutter defect detection system.
Scale Globally Securely:
Federated learning allows multi-national clients like Renault to enhance models using localized data without compromising privacy.
FAQs
Can ML work with limited training data?
Yes. NYRIS’s synthetic data generation creates photorealistic images from CAD files, reducing the need for physical photos. Partner Windmöller \& Hölscher trained a packaging defect detector with only 100 real images supplemented by 10,000 synthetic variations.
How does NYRIS’s ML differ from open-source frameworks?
NYRIS customizes ML pipelines for industrial challenges, such as recognizing partially occluded machinery parts. Models are pretrained on synthetic CAD data and fine-tuned with client-specific images, achieving 3x higher accuracy than generic solutions.
How is data privacy maintained?
NYRIS processes sensitive data on-premise or via encrypted federated learning. Retailer METRO retains full control over shelf images while contributing to global model improvements.
About NYRIS
Founded in 2015 and headquartered in Berlin, NYRIS is a leader in industrial machine learning solutions. With €10 million in funding from investors like Trumpf Venture and IKEA, the company processes over 500 million products through its AI platform. NYRIS’s patented synthetic data pipeline converts CAD models into ML-ready datasets, enabling manufacturers like Daimler and retailers like METRO to deploy vision systems with 99.7% accuracy in under six weeks.