Artificial Intelligence
AI-Assisted Defect Detection
Our on-device neural inference engine processes every frame in real time, surfacing PACP-codeable defects before the crawler exits the pipe.
TXVision Neural Engine v3
Trained on 2.8 million annotated pipe frames spanning concrete, PVC, clay, cast iron, and HDPE substrate types. Achieves NASSCO defect detection accuracy above 94% on unseen validation sets.
Runs on an NVIDIA Jetson Orin NX edge module inside the field control unit — no cloud dependency, zero-latency grading in the field.
94.7%
Defect detection accuracy
2.8M
Training frames
18
Defect classes recognised
<8ms
Per-frame inference latency