Episode Details

Construction sites change daily — different trucks, new equipment, shifting shadows. A layperson sees noise, an expert sees process, but what can AI see that both miss? This episode explores the rapidly maturing field of visual change detection, from Siamese networks to transformer-based architectures like ChangeFormer and DINOv2. We break down how foundation models have made ground-level change detection practical for hyperlocal use cases, the tradeoffs between sensitivity and specificity, and the commercial and open-source tools available today. If you've ever wondered how to build a system that tells the difference between "same truck, different Tuesday" and "different truck entirely," this episode is for you.