KAD3 VisionPick AI
Product recognition and cobot picking using advanced artificial vision.
Product recognition, gripping point estimation and cobot picking
Main capabilities
- Visual recognition of objects via 2D/3D images.
- Localisation and classification of the product to be picked.
- Assessment of the usable surface and gripping point.
- Sending coordinates and parameters to the cobot control system.
- Support for automatic pick&place in warehouse and logistics scenarios.
Operational value
VisionPick AI addresses one of the most delicate points in automation: turning a real scene into a reliable robotic action. The combination of artificial vision and gripping logic reduces picking uncertainty, especially when products have different shapes, orientations or surfaces.
An industrial perception system oriented towards action
Accurate recognition
The system processes images to identify the product and determine the parameters useful for picking.
Grip selection
Surface analysis helps define the gripping mode most consistent with the object's geometry.
Cobot integration
The processed information is transferred to the control layer to make the robotic operation executable.
From visual recognition to picking instruction
Scene acquisition
The camera detects the area of interest and acquires the information needed for processing.
Image analysis
Algorithms identify the product, estimating its position, orientation and characteristics relevant to gripping.
Target calculation
The system determines coordinates and gripping parameters, including criteria for choosing the suction or contact mode.
Picking execution
Information is sent to the control module, enabling the cobot to perform picking and placement in the automatic cycle.
Areas where AI vision supports automatic picking
Frequently asked questions
Does VisionPick AI replace the cobot control module?
No. Its primary role is to generate information useful for picking. Maximum value is achieved when it operates together with a control module and a process orchestrator.
Can it be adapted to different products?
The solution was designed for scenarios with different products and can be customised based on the dataset, sensor and required accuracy level.
Is it useful for product identification alone?
Yes. Although oriented towards pick&place, the platform can also offer value in scenarios where visual recognition is the first step towards broader automation.
Make automatic picking smarter
The KAD3 team can analyse the application context, product type and gripping constraints to define a configuration consistent with customer objectives.
Request a vision and picking demo