From 6 to 8 October, Phaseform exhibited in Start-up World at VISION 2026 in Stuttgart, the world’s leading trade fair for machine vision. Thank you to everyone who came by. And we are very glad that on Tuesday 6 October, our CTO Pouya Rajaeipour’s talk earned Phaseform the title VISION Start-up of the day.
VISION is where the machine vision industry meets: camera and sensor manufacturers, lens makers, illumination specialists, system integrators and the OEMs who turn all of it into inspection, robotics and quality-control systems. For us the show marked a deliberate step — taking adaptive optics out of the research microscope and into industrial imaging.
What visitors saw over three days
- Live wavefront correction: a compact aberration-correction demonstrator on a single USB-C cable, driven from a smartphone app, with a wavefront sensor on the laptop showing each correction as it is set.
- Fovea Imaging, shown by Shi Mao of KAUST: a simple lens and a Delta 7 Deformable Phase Plate under machine-learning control, giving one image that is sharp from center to edge. The method is described in the Fovea Stacking paper, Best Paper at SIGGRAPH Asia 2025.
For machine and robot vision, that means simpler optics — fewer lens elements, looser tolerances — and less data to move and process, when paired with a sensor that reads out only the region of interest at full resolution.
Two talks
On the Industrial VISION Days stage on Tuesday 6 October, our CTO Pouya Rajaeipour presented Algorithm-Driven Optics: 63 Degrees of Freedom for AI-Controlled Imaging in the start-up session — the talk that won us VISION Start-up of the day.
At the EPIC Technology Meeting on Computer Vision Systems and Applications, supported by PhotonHub, our CEO Stefan Weber spoke on From Glass to Algorithms: High-Order Wavefront Correction for Compact Vision Systems — correction that acts before the light hits the sensor.
Why machine vision, and why now
The show made the direction of the industry plain: cameras are moving to 100 and 200 GigE and ever higher resolutions, and AI is taking over a growing share of inspection software. Once the sensor and the data link no longer limit the system, the optics do: the lens decides how much detail reaches the sensor — and an AI model can only learn a defect that the image actually shows.
Optical aberrations are a hidden cost driver in imaging systems. They force additional lens elements, tighter assembly tolerances and expensive active alignment — and they cost yield. Worse, once a system is assembled, a fixed lens cannot adapt: not to a residual manufacturing error, not to thermal drift, not to a new working distance.
The Deformable Phase Plate (DPP) changes that premise. It is a thin, fully transmissive wavefront modulator with 63 electrostatically controlled degrees of freedom — a programmable optical surface that corrects and shapes higher-order aberrations directly in the transmission path. It is polarization-independent, low-power, and thin enough to sit inside a compact objective.
A high actuator count is exactly what makes the DPP a natural counterpart for machine learning: a trained model sets all 63 electrodes at once, for calibration-based, sensorless correction while the system is running.
For OEMs, the interesting conversation is about integration: relaxing tolerances, reducing lens count, correcting residual errors after assembly to raise production yield, and adding dynamic optical functionality under software control. If you design imaging systems and any of that sounds like a problem you recognize, we should talk.
The conversations
The best part was the conversations: imaging systems where fixed optics have run out of room. To everyone we spoke with in Stuttgart — we will be in touch.
Missed us at VISION? Send us your hardest optics problem through our contact page.
