Introducing DEPZ: UHPlast Ltd. R&D Unit for Robotics & Computer Vision
Today we're announcing DEPZ—a new R&D unit at UHPlast Ltd. in Poland, focused on robotics and industrial automation. Our first mission: bring computer vision to the production floor, starting with automated defect detection for large-format products.
The Challenge That Created DEPZ
As production volumes grow and product variety expands, manual quality inspection hits its limits. Consistency drops. Defects slip through. The cost of catching problems late in the process adds up fast.
UHPlast created DEPZ with a clear mission:
- Build engineering solutions that work in real factory conditions
- Connect R&D directly to measurable operational outcomes
- Develop reusable technologies across multiple production scenarios
Why Computer Vision First
Computer vision is one of the highest-impact technologies for industrial automation. It enables fast, repeatable inspection while creating a complete digital record of every product that leaves the line.
For large-format polymer products, defects present unique challenges:
- Subtle surface issues—micro-dents, scratches, faint marks invisible to tired eyes
- Geometry problems—warpage, edge deformation, thickness variations
- Process-dependent flaws—tied to temperature, cooling rates, mold condition
Reliable detection requires more than pointing a camera at a conveyor. It demands controlled capture, robust algorithms, and tight integration with production systems.
Our First System: Post-Line Defect Detection
We're starting with post-line inspection—scanning products after they leave the main production cycle. No disruption to throughput, full visibility into quality.
The system architecture we're building:
- Inspection Cell. Controlled lighting and vibration-isolated mounting eliminate glare, shadows, and noise.
- Capture Pipeline. Synchronized triggering with full metadata—batch ID, timestamp, station, operator mode.
- Detection Engine. Algorithms tuned for large surfaces and production-specific textures.
- Visual Reporting. Defect heatmaps, zoomed evidence frames, and structured QA reports.
- Traceability. Every defect linked to production context for root-cause analysis.
Building the Foundation: Defect Knowledge Base
Beyond software, we're building something equally valuable—a structured defect knowledge base:
- Curated images of real defects and acceptable variations
- Consistent labeling with clear defect taxonomy
- Links to process parameters for pattern discovery
This dataset is the foundation for continuous improvement—enabling us to expand detection capabilities to new product families over time.
What's Next
Our roadmap for the coming months:
- Validate system robustness under real conditions—dust, reflections, surface variation
- Align acceptance criteria with quality specifications
- Move from pilot to production-floor deployment