Perception studio · Aarhus
Vision and control
for machines that act on what they see.
We design perception and control for machines that must interpret the real world in milliseconds, on factory floors, mobile robots, and field hardware.
Trusted by
In the field
Where perception runs
From line-side inspection to mobile platforms and precision integration, the environments our vision and control systems are built to operate in.
Capabilities
What we build
Four ways we work, from a first feasibility read to systems running unattended on the floor.
Computer-vision systems
Detection, tracking, and 3D understanding tuned to your optics, lighting, and latency budget, and proven on your own line.
02Robotics & embodied AI
Perception-to-action stacks for manipulation, navigation, and collaborative workcells, with state estimation you can reason about.
03Edge deployment & MLOps
Packaging, telemetry, and regression hooks from Jetson to industrial PCs, so your team trusts what the fleet reports once it is running on real sites.
04Perception audits & prototyping
An honest read on whether an idea is buildable inside your constraints, before anyone commits to a full programme.
Selected work
Perception, shipped to production
How we work
About the studioPerception is rarely the hard part. Making it reliable enough to leave running, unattended, is.
- Measure, then claim We quantify before we describe. Latency, accuracy, false-reject rates, drift over time. Numbers you can check, taken on your data and your hardware, not a vendor's slide.
- Honest about uncertainty Every perception system has an operating envelope and a failure mode. We tell you where ours sits and where it breaks, so you can plan around the edges instead of discovering them in production.
- Built to ship Research that never leaves a notebook is not finished work. We design for the target hardware from the first week, packaging, telemetry, and regression tests included, so the system survives contact with the real world.
- Restraint The simplest design that meets the spec is the one we keep. We add a sensor, a model, or a dependency only when it earns its cost in reliability or maintenance.
Bring us the hard part
Tell us what the machine needs to see.
Share your environment, sensors, and what “good” looks like. We’ll tell you what is buildable, what is not, and where we would start.
[email protected]