ML project · Fundamentals of Machine LearningFind road damage, square by square
Upload a road photo. The model checks every square, names the damage it sees, and gives the road a repair priority.
The pipeline
- PhotoA phone or dashcam photo of the road.
- GridCut into square cells, each checked on its own.
- PreprocessGrayscale, resized to 32 × 32 = 1,024 numbers.
- FeaturesPCA plus 5 measurements: brightness, contrast, edge strength, dark area, edge direction.
- ClassifySoftmax regression, SVM and a neural network pick one of 5 classes.
- PriorityThe damage adds up to a 0–10 repair priority.
Check a road
Photo uploadArrives once the model is trained.