The theme of this release is self-sufficiency. Model and refine parts inside the software, pick trickier objects, and get commissioning right the first time.

In July 2026, Photoneo released the latest major versions of Bin Picking Studio and Locator Studio. These releases introduce direct creation of 3D localization models from sensor scans, native tools for refining 3D model geometry, and key advancements in AI-based localization like object classification.

Additionally, this update unlocks the rapid Fast-scan feature for CAD MultiView configurations and delivers a host of general usability and performance improvements to the platform.

If you’ve ever stalled a deployment waiting on a CAD file, fought a mesh in third-party software, or re-run a calibration because the settings weren’t quite right, this update is aims at you. Everything below moves one more step of that work inside the app.

What’s new

• In-app 3D model creation
• Mesh-editing tools
• On-demand object classification
• AI-optimized scanning profile
• Calibration scanning profile
• Fast-Scan in MultiView
• Estun robot support

3D model creation from scans

Bin Picking Studio 1.12 builds a model of your target workpiece directly from a 3D scan, right in the web interface, with no CAD required. The new mesh-editing tools let you refine that model in the same place: crop the scene, simplify geometry, remove background noise, reorient normals, and define the origin point.

Until before, a missing or unusable part model was one of the most common bottlenecks in a bin picking deployment. CAD data isn’t always available, and when it is, it’s often too heavy or too detailed to localize against efficiently. The workaround meant exporting scans, cleaning them up in third-party mesh software, and importing the result.

That whole detour now happens in the browser. Simpler, cleaner models are faster to produce, and they localize faster and more accurately in production. The workflow is best suited to semi-structured setups with consistent viewing angles, and the release also adds PLY support alongside STL for imported models.

Pick smarter by knowing more about each object

On-demand object classification lets every object the system identifies carry extra metadata when you ask for it: whether it’s occluded or clear, or which color group it belongs to. Classification is available with AI-based localization.

Previously, a pose estimate told the robot where an object was, but not whether it was actually a good pick, or which of several similar-looking items it was. That gap led to failed grasps on half-buried parts, and it forced you into a separate vision system whenever you needed to sort by type or color.

Now you can prioritize clean, pickable objects and skip the buried ones, or route items by color, all inside the same localization workflow you’re already running.

Better AI recognition, straight from a better image

The new scanning profile for AI recognition is tuned specifically to capture the best possible color image for neural-network inference.

That matters because AI-based localization is only as good as the image it’s given. Weak color images make the neural network work harder and slip more often. And it’s most error-prone on exactly the non-uniform items, like packages and retail goods, that AI localization exists to handle.

Cleaner input means more reliable recognition and fewer misses on the varied, messy objects that are hardest to pick.

Get calibration right the first time

The dedicated calibration scanning profile applies settings proven to deliver high calibration accuracy.

Previously, calibration ran on whatever user-defined scan settings happened to be active. Calibration performed with the wrong settings produces results that might look fine until they aren’t, and the fix is usually a full re-run.

It’s a small addition that removes a common, frustrating source of re-work. Calibrate once, trust the numbers, move on.

Keep your speed gains in multi-camera setups

Fast-Scan now works with the MultiView Module.

Fast-Scan already shaved cycle time by confirming that previously located objects hadn’t moved, so the system could pick them again without a full re-localization. Until now it only worked with a single sensor, which meant any setup using multiple sensors gave that speed up entirely.

Setups that rely on multiple sensors to cover blind spots or wide working areas finally get the same cycle-time benefit. Verify what’s still there, skip the redundant scan, and keep throughput up.

More robots supported out of the box

This release adds Estun to Bin Picking Studio’s robot database, which now spans 275+ models across 11+ leading brands, with ready-to-use templates and full path-planning support for 6-axis arms.

Any brand outside that database means custom integration work before the first pick, and every hour spent on integration is an hour not spent on the application itself.

Every robot brand you can drive without custom integration is one less obstacle between you and a deployment. For Estun users, that work is now done.

Also in this release

Bin Picking Studio 1.12 introduces system load monitoring and logging, giving integrators more visibility into the Vision Controller during operation.

Diagnosing a slow or unstable cell used to mean working from the outside in, with little insight into what the controller was actually doing at the time.

Now the data is there when you need it, whether you’re tuning a new deployment or investigating one already in production.

Ge