Why are we here?
About
OASYS
Road condition monitoring in the Philippines is largely manual: inspectors document defects by eye before maintenance decisions are made. It works, but it is slow, labour-intensive and open to subjective judgement.
What it does
A web-based tool that detects road defects from uploaded images and rates their severity using neural perception combined with rule-based reasoning aligned with engineering standards.
Defects detected
Potholes
Bowl-shaped surface failures
Alligator cracks
Interconnected fatigue cracking
Longitudinal cracks
Parallel to the centreline
Transverse cracks
Perpendicular to the centreline
How it works
Ten stages, from photograph to maintenance bulletin. Neural perception handles what the image contains; symbolic reasoning decides what it means.
Upload
An inspector photographs the road following capture guidelines.
Road detection
The road surface is isolated from the surroundings.
Perspective correction
Inverse Perspective Mapping flattens the view for measurement.
Defect detection
Defects are located and classified by type.
Box filtering
Overlapping duplicate detections are removed.
SAM masking
The exact shape of each defect is segmented.
Skeletonization
The crack structure is traced to a centreline.
Physical metrics
Real-world size is computed via Ground Sample Distance.
Severity reasoning
A fuzzy logic engine grades severity against engineering thresholds.
Summary report
A local language model writes a readable maintenance bulletin.
Severity levels
Graded from measured crack width and crack extent against the DPWH Visual Road Condition Assessment Manual and the City Engineer's validated bands — not from how confident the detector was.
Micro cracks under 3 mm (DPWH Narrow), sparse. Routine preventive maintenance.
Open cracks of 3 to 10 mm, or a single pothole up to 0.25 m². Localized patching or sealing.
Wide structural cracks over 10 mm, alligator cracking over continuous sections, or a larger pothole or several. The road is not serviceable.
Key features
Explainable results
Every grade traces to a rule and a measurement, not a score from a black box.
Distress density measurement
Defect area as a share of the road surface, not just per-defect size.
Real-world size estimation
Lengths in metres and areas in square metres, not pixels.
Automated summary reports
A maintenance bulletin generated locally, with no data leaving the device.
How it is evaluated
Detection and severity accuracy
Precision, recall and segmentation overlap against annotated ground truth.
Neurosymbolic vs. neural-only output
Fuzzy reasoning compared against grading by detection confidence alone.
Agreement with human inspectors
Chance-corrected agreement with licensed engineers grading the same images blind.
Who it helps
Government agencies
Objective, standardised condition data for budget planning.
Road engineers
A visible reason for every severity grade assigned.
Local authorities
Faster triage of which repairs to fund first.
OASYS is a decision-support tool that gives engineers objective, repeatable condition data to prioritise against.