PaveSAM – segment anything for pavement distress

Automated pavement monitoring using computer vision can analyze pavement\nconditions more efficiently and accurately than manual methods. Accurate\nsegmentation is essential for quantifying the severity and extent of pavement\ndefects and consequently, the overall condition index used for prioritizing\nrehabilitation and maintenance activities. Deep learning-based segmentation\nmodels are however, often supervised and require pixel-level annotations, which\ncan be costly and time-consuming. While the recent evolution of zero-shot\nsegmentation models can generate pixel-wise labels for unseen classes without\nany training data, they struggle with irregularities of cracks and textured\npavement backgrounds. This research proposes a zero-shot segmentation model,\nPaveSAM, that can segment pavement distresses using bounding box prompts. By\nretraining SAM's mask decoder with just 180 images, pavement distress\nsegmentation is revolutionized, enabling efficient distress segmentation using\nbounding box prompts, a capability not found in current segmentation models.\nThis not only drastically reduces labeling efforts and costs but also showcases\nour model's high performance with minimal input, establishing the pioneering\nuse of SAM in pavement distress segmentation. Furthermore, researchers can use\nexisting open-source pavement distress images annotated with bounding boxes to\ncreate segmentation masks, which increases the availability and diversity of\nsegmentation pavement distress datasets.\n