Load 2D deep learning model

This algorithm loads a 2D deep learning model. Former name 'load convolutional neural network'.

Supported formats are:

Video - Load 2D deep learning model - convolutional neural network

This algorithm is used with the algorithms 'apply 2D deep learning model' or 'apply 2D deep learning object detection / instance segmentation'.

Supported Domains

Execute;

Parameters

Model format

Select the deep learning model format to be loaded. Choose between:

Once the model is loaded, you can inspect it using the feature Project features > Deep learning (CNN) > Convolutional neural networks > Convolutional neural network model info. It contains information about input and output tensor dimensions, relevant for inference (e.g. 'apply 2D deep learning model' algorithm).

Meta graph file

Select existing model meta graph file. Default "{:Workspc.OutputRoot}\model\model.meta"

(Only available for selection of Model format > eCognition model)

Index file

Select existing model index file. Default"{:Workspc.OutputRoot}\model\model.index"

(Only available for selection of Model format > eCognition model)

SavedModel path

Select existing model meta graph file. Default "{:Workspc.OutputRoot}\model\model.meta"

(Only available for selection of Model format > SavedModel)

Tags

Comma separated list of tags to load the model meta graph. Specifies the tag that represents the inference (prediction) graph head node. The default value for the parameter is 'serve'. (Only available for selection of Model format > SavedModel)

ONNX model

Select existing ONNX model. Default "{:Workspc.OutputRoot}\model\model.onnx"

(Only available for selection of Model format > ONNX model)