Try it on real steel defect data

Running live on HuggingFace Spaces. Upload any steel surface image — the AI returns a confidence score, defect class, and routing decision in under 10 seconds. No login. No install.

masteroAI/mastero-ai-defect-detection — HuggingFace Spaces
01

Upload an image

Use any steel surface photo — ideally from the NEU dataset classes: crazing, inclusion, patches, pitted surface, rolled-in scale, or scratches.

02

Run AI analysis

The SegFormer model runs inference and returns a confidence score (0–100%) and defect class prediction across all 6 categories.

03

Read the routing decision

The platform auto-routes: ≥85% = approved, 60–85% = human review, <60% = rejected. Every decision is logged to the audit trail.

Want to see this running on your own manufacturing data?