DaoAI AD-080
Structured light · 77×70mm · 4μmPrecision small-component inspection · ±8mm range.
PLATFORM · DAOAI WORLD
One platformone engine for four lines.
Data labeling · model training · deployment · OTA updates, fully connected — eight years building one vision foundation model that understands the physical world, shared by inspection, robotics, surveillance and content.
SOVEREIGN AI VALUE
Sovereign by design, from day one.
DaoAI World isn't a cloud API — it's a physical-AI foundation you can install in full inside your own facility. Mapped to IDC's three-layer framework, it natively covers data sovereignty (data never leaves), technology sovereignty (models train themselves on your own line data) and operational sovereignty (full autonomy for highly regulated industries).
- Data sovereignty
Training data stays 100% on the customer's premises, with flexible SDK / API / Docker integration.
- Technology sovereignty
Full-stack in-house, from sensors to the world-model inference engine — no dependence on any external LLM API.
- Operational sovereignty
Meets the full autonomy bar required by government, energy and healthcare critical-infrastructure deployments.
- End-to-end platform
Annotate + train + deploy + OTA in one workbench.
- Live on the Inspur AIStore ecosystem
Strategic distributor Unisplendour — ready to scale channels.
- Multi-task coverage
Instance segmentation / keypoints / anomaly detection / OCR / object detection.
- A vertical agent in 30 minutes
Changeover from days to minutes, zero code.
DAOAI WORLD · CONSOLE
Defect Model · v1.3
3C cosmetic defects · deployed on 14 lines99.7%
30sto train a model
<12msper image
1-20samplesAPDT positive-sample start
OUR THESIS
AGI = one general world model + N vertical small models.
Not "a large language model plus skills or agents." The foundation carries one unified representation of physical reality — space, material, light, motion, causality. The model matrix below is the N derived from it: one small, fast model per task family, inferring on site and retraining from what that site sees, while the foundation keeps them all reading the same physical world.
OUR APPROACH
Knowledge lives on the data side, not piled into the weights.
At DaoAI, moving to a new scene does not mean retraining a bigger model. The base model takes a handful of good-part samples and amplifies them, under physical constraints, into enough training data. A 1–20 sample cold start, a model trained in 30 seconds, a vertical derived with zero lines of code — these follow from that approach, not from parameter count.
AI MODEL MATRIX · 11 MODELS
One workbench covering detection, segmentation, recognition and positioning — cold-start from one good image, train in < 1 minute.
Hybrid modelRotated detectionObject detectionInstance segmentationUnsupervised defect segmentationSupervised defect segmentationOCR / textKeypoint detectionPositioningPresence detectionClassification
Ingest→Smart labeling→Dataset→Train→Deploy→Monitor
SDKC++ · C# · Python
DeploySelf-hosted / enterprise-hosted
Data formatsCOCO · VOC · YOLO v5/v8
CollaborationReal-time multi-user labeling · shared workspaces
HARDWARE · In-house 3D Cameras · Eyes for the physical world
3D structured light + HDR — dark and reflective objects imaged clearly in one capture · 77–477mm FOV · 4μm repeatability · full-stack in-house from sensor to world model.
DaoAI AD-470
Structured light · 477×435mm · 40μmLaptop-scale FOV · inline PCBA inspection.
DaoAI AQ-060
Telecentric · 60×60mm · 18MPShadow-free 4-way · PCB solder & placement.
WeRobotics 300VT
3D vision · full systemFeature-less 3D guidance · adaptive.
DaoAI IN-XS
Ultra-high res · 100–140mm5MP snapshot inspection, 5μm Z repeatability, for consumer-electronics parts.
DaoAI IN-S
Small parts · 150–270mm5MP snapshot inspection with a 132 × 106mm field of view.
DaoAI IN-M
Automotive parts · 270–450mm5MP snapshot inspection, 225 × 185mm field of view, up to 450mm.
See the 3D camera product page →
INSPECTION CAMERAS · SPECS
Six inspection models, from a 60 × 60mm telecentric field at 2μm repeatability to a 477 × 435mm large-format field — one optical stack, one point-cloud pipeline.
| Specification | AQ-060 | AD-080 | AD-470 | IN-XS | IN-S | IN-M |
|---|---|---|---|---|---|---|
| Working distance (mm) | 32 | 186 | 510 | 100–140 | 150–270 | 270–450 |
| Field of view (mm) | 60 × 60 | 77 × 70 | 477 × 435 | 83 × 60 | 132 × 106 | 225 × 185 |
| Measuring range | ±5 mm | ±8 mm | ±40 mm | — | — | — |
| Image resolution | 18 MP · 4288 × 4288 | 18 MP · 4496 × 4096 | 18 MP · 4496 × 4096 | 5 MP | 5 MP | 5 MP |
| Repeatability | 2 μm | 4 μm | 40 μm | 5 μm | 8 μm | 15 μm |
| Calibration accuracy | — | — | — | 20 μm | 25 μm | 30 μm |
| Spatial resolution (mm) | — | — | — | 0.03 | 0.05 | 0.09 |
| 3D acquisition time (s) | 1 | — | — | 2 | 2 | 2 |
| Illumination / projection | 4-way structured light · RGBW ring · telecentric lens | Dual projector · white LED | Dual projector · white LED | Dual-cam · blue LED | Dual-cam · blue LED | Dual-cam · blue LED |
| Data interface | CoaXPress 2.0 (CXP-12×4) | 10 GigE | 10 GigE | Ethernet | Ethernet | Ethernet |
| Power | 24V DC 12A | 24V DC 10A | 24V DC 10A | — | — | — |
| Dimensions (mm) | — | — | — | 224 × 286 × 98 | 208 × 286 × 98 | 208 × 286 × 98 |
| Weight (kg) | — | — | — | 3.8 | 3.7 | 3.7 |
← swipe to see the full spec table →
Shared specs · 0–40°C operating range · passive cooling · IP65 industrial housing · 3D HDR captures dark and specular surfaces clearly in a single shot · contrast / outlier / cluster / Gaussian / hole-filling point-cloud filters · on-site calibration restores accuracy in under three minutes · IN series carries NVIDIA® Jetson Xavier™ NX.