EasySegment - Detecting and Segmenting Defects

EasySegment is the deep learning segmentation library of eVision.

It contains 2 different modes:

• The unsupervised mode:

  • The tool is trained only with good images (EUnsupervisedSegmenter ^{.NET Python} class).

  • This mode does not require a ground truth segmentation and the creation of the dataset is thus much quicker than for the supervised mode.

  • This mode can detect unexpected defects while the supervised mode is only capable of detecting defects similar to those in the dataset.

• The supervised mode:

  • The tool is trained using the ground truth segmentation defined for the images (ESupervisedSegmenter ^{.NET Python} class).

  • This mode can detect and segment more types of defects with better accuracy than the unsupervised mode. It directly builds a model for the defects while the unsupervised mode builds a model of the good images and tries to detect variations from this model.

  • You can also use this mode to segment other types of objects than defects.