30 Million Picks Later, Eureka Robotics Lets Anyone Set Up 3D Bin-Picking Without Code

Eureka Robotics launched StereoPick, letting engineers set up a 3D bin-picking system straight from a CAD file in six steps, no code or vision expertise required.

Key Takeaways

  • Eureka Robotics launched StereoPick, letting engineers set up 3D bin-picking from a CAD file in six steps without code or manual image annotation.
  • The software builds on an Eureka AI Vision System that has logged more than 30 million production picks to date.
  • Simple, rigid parts can reportedly be trained within hours, down from the days or weeks a custom vision integration typically requires.
  • StereoPick runs on local hardware with no cloud dependency, works with any robotic arm, and carries no subscription fee.
  • Eureka's vision system is already deployed at Pratt & Whitney, Subaru and Coherent.

Eureka Robotics has launched StereoPick, software that lets factory engineers set up a 3D bin-picking system straight from a CAD file, without writing code, hiring a vision specialist, or manually annotating training images. The Atlanta-based company announced the product this week, building on an Eureka AI Vision System that it says has already logged more than 30 million picks in production.

The pitch is speed and independence from integrators. StereoPick converts a part's CAD model into synthetic training images, validates the setup against a digital twin of the real cell, and walks an engineer through six guided steps to get a production-ready pick running. Eureka says simple, rigid parts can be trained within hours rather than the days or weeks a custom vision integration project typically takes, according to RoboticsTomorrow.

The software runs locally rather than depending on a cloud connection, pairs with any robotic arm rather than locking customers into a specific brand, and carries no subscription fee. Eureka is already running the underlying vision system at Pratt & Whitney, Subaru and Coherent, giving the launch a base of industrial references before a single new customer signs on.

Bin picking's stubborn bottleneck

Picking parts out of a bin is one of the oldest unsolved problems in industrial robotics. Parts arrive jumbled, overlapping and inconsistently lit, and traditional 2D vision systems choke on the ambiguity. 3D vision solved the accuracy problem years ago, but setup remained the bottleneck: configuring a 3D picking cell has usually meant hiring a systems integrator, annotating thousands of real images by hand, and spending weeks tuning the result before a single part gets picked reliably. That bottleneck has slowed automation in exactly the industries — aerospace, automotive and electronics — that most need relief from bin-picking labor shortages.

StereoPick attacks that setup cost directly by using CAD data instead of hand-labeled photos, a shortcut that only works because Eureka already has 30 million real-world picks' worth of data informing how its synthetic training pipeline generalizes to messy, real factory bins. CEO Cuong Pham framed the goal as putting the customer's own team in control of the system from setup to production, rather than leaving them dependent on the vendor every time a new part needs to be added to the line.

Where it sits against rival picking systems

Eureka is entering a market where ABB, Fanuc and a wave of startups all sell vision-guided picking in some form, and where ABB alone has recently touted AI-driven setup-time cuts of up to 90% for its own vision trainer. The common thread across these launches is the same: picking accuracy is no longer the hard problem, onboarding speed is. Vendors are racing to compress the gap between buying a 3D camera and having it actually pick parts on a live line.

Eureka's bet is that an open platform with no subscription fee and no cloud dependency will appeal to manufacturers wary of vendor lock-in, especially smaller suppliers that cannot justify a dedicated integrator relationship for every new part number. Whether that undercuts rivals on price enough to matter will depend on details Eureka has not yet disclosed, including how StereoPick is priced relative to the hardware bundle it ships with.

The company has not said how many customers have adopted StereoPick since launch or whether its existing manufacturing partners are already running it. The clearest signal of whether the no-integrator pitch works will be how many of Eureka's next customers skip hiring outside vision help entirely.

Category: Industrial Robots

Tags: Industrial Robots Bin Picking machine vision Eureka Robotics StereoPick CAD-based training

Related Articles

Frequently Asked Questions

What is StereoPick?

StereoPick is software from Eureka Robotics that lets manufacturers set up a 3D bin-picking system directly from a CAD file, without coding, vision expertise or manual image annotation.

How long does it take to train StereoPick on a new part?

Eureka Robotics says simple, rigid parts can be trained within hours, compared with the days or weeks a traditional vision integration project typically takes.

Does StereoPick require a subscription or a specific robot brand?

No. Eureka says the platform runs on local hardware without a cloud connection, carries no subscription fee, and integrates with any robotic arm.

Who is already using Eureka Robotics' vision system?

The underlying Eureka AI Vision System is deployed at Pratt & Whitney, Subaru and Coherent, and has logged more than 30 million picks in production.