State-of-the-art approaches to detecting ridge-like structures in images rely on filters designed to respond to locally linear intensity features. While these approaches may be optimal for ridges whose appearance is close to being ideal, their performance degrades quickly in the presence of structured noise that corrupts the image signal, potentially to the point where it truly does not conform to the ideal model anymore. In this paper, we address this issue by introducing a learning framework that relies on rich, local, rotationally invariant image descriptors and demonstrate that we can outperform state-of-the-art ridge detectors in many different kinds of imagery. More specifically, our framework yields superior performance for the detection of blood vessel in retinal scans, dendrites in bright-field and confocal microscopy image-stacks, and streets in satellite imagery.