3D Vision that Sees in Depth: RGB-D cameras and 3D point clouds identify the position and shape of objects in three dimensions.
PRODUCT
AI-3D Vision Robot
Giving robots “eyes” and “judgment” — 3D vision and AI recognize randomly mixed, irregular objects for precise picking.
Conventional robots move only according to predefined coordinates, so even small changes in an object's position or shape can cause errors or require repeated teaching. STIGMA's AI-3D Vision Integrated Robot recognizes objects in three dimensions with a 3D camera, while AI independently calculates where and how to grasp them. It precisely picks randomly mixed, irregular objects through bin-picking and inspects their surface condition at the same time.
How It Works
AI Grasping that Decides on Its Own: AI separates each object through instance segmentation and automatically calculates the most stable grasp point and angle (6D pose).
AI Quality Inspection During Picking: AI learns and filters even undefined defects such as surface contamination, loose yarns and shape distortion.
Virtual Verification Before Installation (Digital Twin): Robot motions are verified in a virtual space before installation, reducing on-site trial and error and deployment risks.
What Makes It Different from Conventional Robots?
| Conventional Robot | STIGMA AI-3D Vision Robot | |
|---|---|---|
| Control Method | Moves only to fixed coordinates → Positional deviation causes errors or collisions | Vision corrects errors in real time → Precision within ±1 mm even in irregular environments |
| Quality Control | Performs transfer only; inspection requires separate personnel and equipment | Inspects while grasping → Automatic sorting reduces defects |
| Scalability | Fixed use for products of a specific size | Handles various sizes and items by changing the gripper and settings |
What changes after deployment?
- 01
Less need to manually align positions or reteach the robot.
- 02
The robot judges and works independently even as site conditions change.
- 03
Higher process precision and productivity with fewer defects and safety incidents.
Deployment & Verification Cases
AI Winding-Roll Prediction and Inspection Integrated Robot System
Stacking error within ±1 mmTarget: 20% defect reduction
The robot uses 3D vision to recognize, grasp, transfer and precisely stack yarn winding rolls. AI vision inspects surface defects during stacking, while a digital twin monitors the process in real time. — Verification in progress in 2026
Integrated Nonwoven-Roll Winding, Transfer, Palletizing and Packaging Automation — MOTREX ADM
Productivity +125.5%Defect rate -62.5%Cost -50%
Winding, transferring and stacking heavy nonwoven rolls are integrated into one automated process using a multi-gripper and digital twin. — Verification in progress in 2026; installation scheduled for completion in October
The 3D vision and AI grasping technology was developed through the Ministry of SMEs and Startups' Startup Growth Technology Development Program (Didimdol), “Development of a 3D Vision Bin-Picking Robot and Gripper” (KOROS performance evaluation and technical-data escrow completed), and is being enhanced with an NVIDIA Omniverse Isaac Sim-based digital twin.