RGB + pose
Axes on every detected part, ready for approach.
SEDECO
3D bin picking built to integrate
Affordable industrial 3D sensing, AI pose, grasp definition and robot integration, for applications where premium platforms are commercially hard to justify. Open, configurable, suited to accessible robot platforms. Send an STL or samples; we start a feasibility check.
Scroll · scan
Live stack
RGB, depth and pose, in one view.
Our 3D AI inference sees the bin, fits objects and stages the best grasp for the robot. Camera, integrations and outputs online.
Metric depth
Aligned depth map for volume and occlusion.
Robot online
Controller linked, grasp to TCP.
Depth + metrics
See what the camera measures.
Aligned metric depth beside live RGB. Capture, infer and drop timings stay visible, so you know what the stack does on the line.
Grasp tooling
Pick-point on the model, not guesswork.
Choose a point on the detected STL. Position follows every pose; distance runs along the surface normal. Approach direction and cylinder axis stabilize tool orientation.
- Relative distance along approach
- Cylindrical axis for symmetric parts
- Straight to robot TCP
Detect → control
Score objects. Wire the robot.
Cell
Hardware that finishes the grasp.
Fin Ray-style or mechanical fingers, same vision stack, different end effector. Dirty parts, occlusion, pose variation: that is why 3D.
Pipeline
From STL and point cloud to robot motion.
In-house tooling for training, synthetic data, recursive improvement and grasp selection, wired to the robot.
01
Load STL
CAD as ground truth for fit and grasp.
02
Point cloud
Multi-GigE or Hikrobot 3D, scene to depth.
03
Train the model
Fit under real occlusion and pose variation.
04
Synthetic data
Accelerate learning when labels are scarce.
05
Improve recursively
Feedback until the grasp is dependable.
06
Pickup → robot
Grasp point to controller and PLC handshake.
PoC
STL ready? Samples? Both work.
Workflow: samples or STL → feasibility check → train/configure → robot coupling → commissioning. Ship parts without CAD and constructors can build an STL. Acceptance criteria and support path are agreed up front.
Economically attractive, not universal
We do not position this as a universal replacement for Keyence, Mech-Mind or Pickit. It is a more open, configurable path backed by Sedeco’s machine-vision expertise, performance always subject to feasibility testing on your parts.
Ready to empty your bin?
Share STL or samples, robot brand and takt, we’ll propose a PoC path.
Learn before you buy
Start with the knowledge guides when you need to validate feasibility, cycle time or when 3D depth is actually required for robot vision.