You are here: Start » Machine Vision Guide » Deep Learning

Deep Learning

Overview

Deep Learning method enables the user to analyze images containing complex computer vision problems. Deep learning is a milestone in computer vision, as it helps solve problems that are too difficult for classic computer vision algorithms.

Defects found in textile material.

Main advantages of this method:

  • Simple configuration,
  • high performance,
  • flexibility of use.

Adaptive Vision offers two major algorithms based on Deep Learning:

  • Anomalies Detection - detection of differences based on the images provided for training,
  • Feature classification (segmentation) - extraction of image parts based on patterns learned from image training set marked by the user.

Main applications of Deep Learning:

  • Detection of surface defects: cracks, damages or discoloration,
  • detection of shape defects: superfluous elements, missing parts, shape deformations,
  • advanced edge detection,
  • finding defects in object patterns,
  • quality analysis of objects without fixed shape.

Road pattern extraction from a satellite image.

Anomalies detection

This technique is used to detect samples with any potential defects, deformations, anomalies or damage. It only needs a set of fault-free samples to learn the normal appearance and several faulty ones to define the level of tolerable variations.

Example of shape deformation detection.

Feature classification

This filter is used to precisely segment one or more classes of features or objects within an image. The pixels belonging to each class must be marked by the user in the first step. The result of this technique is a list of probability maps for every class.

Example of veins extraction from a retina image.

See also: