Problem: Trainieren Eines Deep-Learning-Modells – "Modelltyp Wurde Nicht Gefunden" Oder Fe

Beschreibung

When attempting to Train a Deep Learning Model it is possible to add image chips but not be presented with the desired model type.
If the desired model type is selected first, and the image chips are then added, Error 00800 may be returned.

Ursache

The cause of the model type not being found and Error 00800 is an incompatible Meta Data Format selected when exporting training data with the Label Objects for the Deep Learning tool. The Meta Data Format dictates what model types the user can use and/or see when Training Deep Learning Models.

Lösung oder Problemumgehung

To resolve this issue it is necessary to change the Meta Data Format to one compatible with the required model type. The below tables show which model types are compatible with Meta Data Format types.

Classify Pixels Using Deep Learning
	Meta Data Format Type
Model Type		DeepLabV3	U-Net	The Pyramid Scene Parsing Network	BDCN Edge Detector	HED Edge Detector	Multi Task Road Extractor	ConnectNet	Change Detector
	KITTI Labels	No	No	No	No	No	No	No	No
	PASCAL Visual Object Classes	No	No	No	No	No	No	No	No
	Classified Tiles	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
	RCNN Masks	No	No	No	No	No	No	No	No
	Labeled Tiles	No	No	No	No	No	No	No	No
	Export Tiles	No	No	No	No	No	No	No	No
	CycleGAN	No	No	No	No	No	No	No	No
	Multi-labeled Tiles	No	No	No	No	No	No	No	No

Classify Objects Using Deep Learning
	Meta Data Format Type
Model Type		Feature Classifier
	KITTI Labels	No
	PASCAL Visual Object Classes	No
	Classified Tiles	No
	RCNN Masks	No
	Labeled Tiles	Yes
	Export Tiles	No
	CycleGAN	No
	Multi-labeled Tiles	Yes

Detect Objects Using Deep Learning
	Meta Data Format Type
Model Type		Single Shot Detector	RetinaNet	MaskRCNN	YOLOv3	FasterRCNN
	KITTI Labels	Yes	Yes	No	Yes	Yes
	PASCAL Visual Object Classes	Yes	Yes	No	Yes	Yes
	Classified Tiles	No	Yes	No	Yes	Yes
	RCNN Masks	No	No	Yes	No	No
	Labeled Tiles	No	No	No	No	No
	Export Tiles	No	No	No	No	No
	CycleGAN	No	No	No	No	No
	Multi-labeled Tiles	No	No	No	No	No

Detect Change Using Deep Learning
	Meta Data Format Type
Model Type		Image Captioner	Super-Resolution	CycleGAN	Pix2Pix
	KITTI Labels	No	No	No	No
	PASCAL Visual Object Classes	No	No	No	No
	Classified Tiles	No	No	No	No
	RCNN Masks	No	No	No	No
	Labeled Tiles	No	No	No	No
	Export Tiles	Yes	Yes	No	Yes
	CycleGAN	No	No	Yes	No
	Multi-labeled Tiles	No	No	No	No