Healthcare
Earth Observation
Industry 5.0
Waste Management
Optical Sorting
Classifying 20+ Material Types to Improve Optical Waste Sorting
December 23, 2025
Pellenc ST
Pellenc ST develops advanced optical sorting solutions for waste management, providing real-time monitoring of recyclable product quality and waste stream purity.
Developing high-accuracy AI models capable of recognizing over 20 classes of materials in real-time, including complex PET variations and multi-layer contaminants that emit identical infrared signals.
Project Goals
- Develop annotation protocols for multiple waste streams
- Enable precise waste classification from optical imagery
- Improve real-time monitoring of waste quality
Introduction
Pellenc ST develops advanced optical sorting solutions for waste management, providing real-time monitoring of recyclable product quality and waste stream purity.
Main Challenge
Developing high-accuracy AI models capable of recognizing over 20 classes of materials in real-time, including complex PET variations and multi-layer contaminants that emit identical infrared signals.
Goals
- Develop annotation protocols for multiple waste streams
- Enable precise waste classification from optical imagery
- Improve real-time monitoring of waste quality
Precision AI for High-Throughput Circular Economy
Redefining Purity Standards in Global Waste Management
Project Impact
50,000+
Massive Data Scale: Annotated 50,000+ images since July 2023, supporting continuous model evolution.
22
Operational Efficiency: Achieved an average annotation speed of 22 seconds per polygon, essential for rapid model retraining.
Enhanced Sorting Accuracy: Improved the detection of metals and recoverable cellulosic waste within complex plastic streams.
Multi-Material Stream Enrichment
High-Accuracy Polygon Segmentation
Delivered precise polygon-level annotation to capture the exact boundaries of recyclable materials and contaminants, ensuring high fidelity for AI model training.
0
1
High-Accuracy Polygon Segmentation
Custom Annotation Protocols for 3 Critical Waste Streams
Developed structured annotation guidelines tailored to three key waste streams, enabling consistent labelling across diverse material types.
0
2
Custom Annotation Protocols for 3 Critical Waste Streams
Contaminant Classification and Purity Benchmarking
Applied advanced rules to classify metals, recoverable cellulosic waste, and other contaminants, providing benchmarked datasets for improved sorting accuracy.
0
3
Contaminant Classification and Purity Benchmarking
Real-Time Quality Monitoring Support
Maintained annotation consistency and quality to feed AI-driven real-time monitoring systems for recyclable product purity.
0
4
Real-Time Quality Monitoring Support
20+ Material Classes Including Contaminants
Annotated over 20 distinct material types, including plastics, metals, and other contaminants, creating comprehensive datasets for robust AI sorting.
0
5
20+ Material Classes Including Contaminants
No items found.
