Desion offers AI-driven computer‑vision inspection systems that detect and classify textile defects in real time, integrating directly into existing production lines via standard industrial protocols. The hardware captures multi‑spectral images, while cloud‑hosted analytics provide dashboards, trend reports, and alerts to help manufacturers and laundries reduce rework and material waste.
Funding
Funding not disclosed
Founders
Product
Problem
Textile manufacturers and laundry operators rely on manual visual inspection, which is slow, labor‑intensive, and prone to human error, leading to missed defects, rework, and material waste.
Solution
Desion delivers AI‑driven computer‑vision systems that integrate directly into existing production lines to perform continuous, real‑time inspection of textiles. Trained deep‑learning models automatically detect and classify defects such as holes, stains, doubled or broken yarns, and surface irregularities. The system localizes each defect on high‑resolution images and streams the results to a cloud‑based analytics platform. Operators access a web dashboard that aggregates defect rates, trend analytics, and actionable alerts, enabling rapid process adjustments. Integration is achieved through standard industrial protocols (TCP‑IP, Profinet, Modbus) and optional ERP or FHIR connectors, allowing seamless data flow without line downtime. The modular hardware can be retrofitted to a variety of product formats—from workwear and gloves to rolled fabrics and medical stockings—supporting both standalone and fully automated sorting configurations.
Target Audience
Primary customers are textile manufacturers, industrial laundries, and recycling facilities that process workwear, gloves, medical compression stockings, and rolled fabrics and require automated, high‑precision quality control.
Features
- Multi‑camera, multi‑spectral imaging rig with hyperspectral lighting for robust defect capture on diverse textile surfaces
- Convolutional neural network pipelines that perform pixel‑level segmentation and classification of >30 defect types in under 1 s per item
- Precise defect localization with overlay pictograms for targeted repair or sorting decisions
- Cloud‑hosted analytics engine delivering real‑time KPI dashboards, defect trend reports, and automated quality alerts
- Open industrial interfaces (TCP‑IP, Profinet, Modbus) and optional ERP/FHIR APIs for seamless integration into existing MES systems
- Modular hardware architecture enabling quick retrofits, remote diagnostics, and scalable expansion across product lines
- Configurable quality profiles that can be customized per customer, material type, or regulatory standard