Notes From SCIIF Shenzhen: Machine Vision and AI Sorting Are Catching Up to Our Recovery Ambitions

We spent this month at SCIIF — the South China International Industrial Expo in Shenzhen — and came back with a clear takeaway: the hardware and AI tooling that underpins e-waste recovery is moving faster than most people in our industry realise.

A few things stood out from the floor.

What's Changed Since Last Year

1. Machine vision is genuinely real-time

Polymer type, surface contamination, component density — all classified in real time on belts moving at industrial throughput. Twelve months ago this was demo-grade; today it's deployment-grade. The latency that used to bottleneck sorting decisions is now gone.

2. Robotic sorting arms have caught up

Picking speed and positional precision have both jumped. Multi-axis arms that used to be a compromise between "fast" and "accurate" are now both. Throughput per arm is materially higher than it was at last year's expo.

3. AI classification models compound with data

The vision models on display had been trained on increasingly large, increasingly diverse datasets. The result: better minority-class detection (the rare materials that drive economics), better edge-case handling, and meaningful drops in mis-sort rates.

4. The cost curve has bent

Crucially — and this is the part the headlines miss — the cost of all of this has dropped sharply over the past 12 months. What was a capital-expense conversation in 2025 is increasingly an operating-expense conversation in 2026.

What This Means for Hermion and Our Facilities

Our Hermion AI robotic sorting platform is what drives our recovery rates today — up to 99% precious metal recovery and 75%+ plastic recovery. Those are the numbers we publish, and they hold up against an industry baseline of 85–90% and 40–60% respectively.

But the right way to read those numbers isn't as ceilings — it's as the current floor. The technologies on display at SCIIF feed directly into the next generation of Hermion. We are already planning integration paths into our Preston and Riverbank facilities, and into every planned site that follows.

The economics of urban mining are constrained by sorting accuracy more than any other single factor. As accuracy compounds, the recoverable fraction of every input stream grows — and the line between "waste" and "feedstock" moves further in our favour.

The Honest Question

The expo prompted a question we keep coming back to: which of these capabilities will move the recovery-rate needle the most in the next 24 months?

• Smarter vision models that catch more minority classes?
• Faster mechanical actuation that lets us throughput more material?
• Better material data that improves downstream refining decisions?
• Or the cost curve itself — making advanced sorting affordable at every scale of facility, not just the flagships?

We have our own view, but we are also genuinely interested in where the rest of the industry sees the biggest lever. If you have a perspective — operator, researcher, supplier, or partner — get in touch. The conversation is more interesting when it's not just us talking.

---

Where this fits: 3R Technology Global operates AI-driven recovery facilities across the UK, USA, and Netherlands. Our Reusel engineering centre designs the recycling processes and machinery deployed across the group, including future generations of the Hermion platform.