Edge Extraction with Deep Learning

Deep edge extraction is a novel and unique method for extracting stable edges (e.g., object boundaries) that comes with two main advantages:

For scenarios where various types of edges are observable in an image, deep edge extraction can be trained with just a few images to confidently extract the desired edges. Thus, with MVTec HALCON, the programming effort for extracting specific types of edges is greatly reduced.
Additionally, pre-trained networks are inherently capable of detecting edges in low-contrast and high-noise conditions, enabling the extraction of edges that ordinary edge detection filters cannot detect. In the image below, you can see that edge detection is well performed even in conditions of high noise and the presence of texture in the background, and so on.