icon

Sewer Pipeline Defect Detection

Understand the true state of your sewer infrastructure easily by quickly analyzing CCTV data and detecting sewer defects found.

100 credits (with a subscription)

200 credits (without a subscription)

Version 1.0
Free Trial available!
Sewer Pipeline Defect Detection
Powerful Computer Vision
This machine learning model uses object detection to detect and classify structural and operational defects as well as construction features in CCTV data, using NASSCO standards.
Use Case
Designed to streamline the quality control process for busy consultants, this model eliminates the need to spend hours manually viewing coded CCTV data.
Fast Analysis
The model is currently trained to detect the following defects/features:
  • Access Point Manholes (AMH)
  • Broken pipes (B)
  • Cracks, including:
    • Circumferential Cracks (CC)
    • Multiple Cracks (CM)
  • Deposits, including:
    • Deposits Attached Encrustation (DAE)
    • Deposits Attached Grease (DAGS)
    • Deposits Attached Other (DAZ)
    • Deposits Settled Compacted (DSC)
  • Fractures, including:
    • Circumferential Fractures (FC)
    • Longitudinal Fractures (FL)
    • Multiple Fractures (FM)
  • Infiltration Stain Joint (ISJ)
  • Roots, including:
    • Roots Ball Joint (RBJ)
    • Roots Fine Joint (RFJ)
    • Roots Medium Joint (RMJ)
    • Roots Tap Joint (RTJ)
  • Taps, including
    • Tap Break-in / Hammer Defective (TBD)
    • Tap Break-In Defective (TBI)
    • Tap Factory (TF)
    • Tap Factory Activity (TFA)
    • Tap Factory Capped (TFC)

Required Inputs

  • PNG, JPG, or JPEG files

Expected Outputs

  • Annotated images
  • A summary report in XLSX format

The Sewer Pipeline Defect Detection model analyzes CCTV imagery and identifies/classifies a wide variety of defects found, based on NASSCO standards. Designed to increase inspection speed and provide additional quality control to the classification process, this model is a high-performance solution that frees up time for engineers and contractors, allowing them to focus on higher value activities.

Note: Current mAP (Mean Average Precision) score is 46.5%. Results will continue to be refined as the model is trained on more data.

Cookie Statement

We collect cookies to improve your experience with BRYX.