Playcraft Distress

Turns network survey vehicle, dashcam or phone footage into a geo-referenced distress inventory. It does not merely flag defects — it sizes them, so quantities feed into maintenance estimation rather than a count of red boxes.

Send us 15 minutes of footageSee it run on a call

Deployed with setup support.

Distress detection on survey footage — chainage 0.37 km

A frame of network survey vehicle footage with four pavement defects segmented in amber. Each carries a class label and confidence: three longitudinal cracks at 0.83, 0.67 and 0.37, and an alligator-cracked patch at 0.49. A metadata banner across the top shows chainage 0.37 km and speed 51 km/h; client-identifying fields are covered by solid bars.
Pixel-accurate segmentation, not bounding boxes — because the mask is what the area measurement is computed from. Each defect carries its class and the model’s confidence, and the banner it sits under supplies the chainage that geo-references it. Client-identifying fields are redacted; chainage and speed are not.

What it does

Capabilities

  • Up to 14 distress classes: cracking by type, potholes, ravelling, failed patches
  • Per-defect dimensions, converted to real-world units where camera calibration is supplied
  • GPS and chainage recovered by OCR from the in-video metadata banner
  • Chainage-aware deduplication — one physical defect, one record
  • Region-of-interest masking to exclude bonnet, sky and roadside furniture
  • Review interface for accept, reject or reclassify before a report is final

How it is delivered

Delivery

  • Offline desktop application for Windows
  • Machine-bound licence with per-licence feature gating
  • Optional GPU acceleration; runs on CPU where none is present

What you get back

Outputs

  • Deduplicated defect inventory as CSV — class, confidence, dimensions, chainage and position per defect
  • GeoJSON for direct import into GIS and asset management systems
  • Annotated review video with class labels and confidence
  • Segment-wise summary aggregated along chainage intervals

What you actually receive

The defect inventory

Eleven deduplicated defects across approximately 0.9 km of a real survey — the same pass the frame above came from. One physical defect, one row, regardless of how many frames it appeared in.

Dimensions are reported in pixels here. Converting them to metres needs the camera height and field of view, or the frame width and length, supplied per survey vehicle — a calibration step, not a guess. Uncalibrated footage still ranks severity consistently along a route.

One row carries no chainage. That is a GPS dropout in the source footage, left in rather than tidied away — it is what the file actually contains.

Deduplicated defect inventory

tracks.csv
ClassConf.Length pxWidth pxChainage km
Alligator0.7521256.0374.000.13
Alligator0.8821617.0286.000.26
Longitudinal0.8081369.0172.330.28
Longitudinal0.7281266.0149.820.36
Longitudinal0.7601904.0239.660.47
Longitudinal0.8241797.0198.00— no fix
Alligator0.631871.0234.000.78
Raveling0.592202.353.920.87

Also available

Other tools