Comparative Analysis of ML Models for Street Pothole Detection

Civic infrastructure teams need quantitative guidance on which detectors reliably spot potholes from street imagery before deploying expensive repairs.

I benchmarked CNNs, ResNet-50, and decision-tree baselines in Python/TensorFlow with rigorous dataset splits, validation protocols, and comparative metrics suited to an academic paper.

• Documented trade-offs between deep models and classical learners on the same corpus.
• Public GitHub artifact enabling reviewers to reproduce charts and tables.