Identify and segment surface defects, cracks, scratches, and anomalies on any material at full production speed. Pixel-level precision powered by deep learning — replacing slow, inconsistent manual inspection with always-on AI.
Manual inspection misses up to 30% of surface defects — eyes get tired, lighting changes, defects hide in patterns. MachVizyon's defect detection runs continuously at line speed, applying convolutional neural networks trained on millions of defect samples. From hairline cracks on cold-rolled steel to micro-pits on machined aluminum, the system localizes each anomaly down to the pixel and triggers immediate reject signals through your existing PLC.
Models adapt to your line's specific defect taxonomy. New defect types can be added in days, not months — annotate a few hundred examples and the model retrains automatically.
Every capability is hardened against the noise, vibration, and lighting variability of real factory floors.
Outputs precise defect masks — not just bounding boxes. Quantify size, area, and severity for every anomaly detected.
Distinguishes between cracks, scratches, dents, pits, inclusions, roll marks, and custom defect classes specific to your line.
NVIDIA TensorRT-accelerated models run on edge GPUs, keeping pace with 1,200+ parts/min lines without slowdown.
Edge cases flagged by operators are queued for retraining. Models improve weekly without service interruption.
Each defect gets a calibrated confidence and severity score, letting you set acceptance thresholds per product spec.
Drops onto existing camera mounts. Calibrates to your lighting and conveyor in under a shift, no production interruption.
The same core architecture, tuned for the defect profiles of each sector.
Detects roll marks, scale, inclusions, and edge cracks on hot- and cold-rolled steel coils up to 1,800 m/min strip speeds.
Catches stamping cracks, weld porosity, splatter, and paint defects on body panels and structural welds before downstream assembly.
Identifies particulates, glass cracks, fill-level deviations, and foil tears on vials, ampoules, and blister packs at 800 units/min.
Measures surface finish, detects machining marks, burrs, and dimensional deviations on CNC and forged components.
Spots inclusions, bubbles, scratches, and edge chips on float glass, automotive glazing, and technical ceramics.
Verifies print registration, detects smudges, missing labels, and color-out-of-spec on cartons, labels, and flexible films.
A deterministic pipeline tuned end-to-end for industrial latency budgets.
GigE Vision cameras synchronized to encoder triggers capture each part at controlled exposure.
Frames undergo flat-field correction, ROI cropping, and contrast normalization on the edge GPU.
U-Net-based segmentation model produces a per-pixel defect mask with class probabilities.
Detected regions are classified by defect type and scored by severity against your acceptance rules.
Reject signal fires to PLC over OPC-UA / digital I/O. Result and image archived to MES with audit trail.
Metrics aggregated across active steel, automotive, and pharmaceutical lines.
Hardened for industrial environments. Integrates with the systems you already run.
Send us a sample of your parts or a short clip from your inspection station. We'll show you exactly what AI catches that your current process misses.