Deep learning features

Deep learning features are used in classification based on neural networks (deep learning models).

Deep learning > Number of sample patches

Project features > Deep learning > [Number of sample patches]

Feature that calculates the number of existing sample patches for a class.

Editable Parameters

• Sample folder: Select an existing sample folder path.
• Class: Select the class(es) to calculate the number of sample patches.

Deep learning > Deep learning model info

Project features > Deep learning > Deep learning model info

This feature returns the current settings of the deep learning model. Right-click the feature in the feature view and select Display in Object Information.

The following parameters are displayed for eCognition model:

• Number of layers including hidden layers
• Kernel size for each hidden layer
• Number of feature maps for each hidden layer
• Pooling - max pooling set to yes or no for each hidden layer
• Sample size [Sample patch size x Sample patch size x Number of image layers]
• Classes - output feature class of the image

for TensorFlow SavedModel and ONNX:

• Input tensor information (name, shape, type)
• Output tensor information (name, shape, type)

Learn more:

Deep Learning Tutorial

Deep Learning Classification (User Guide)

Deep Learning Algorithms (Reference Book)

eCognition tv - Deep Learning webinars and more on our website