Object based image classification

[HelenaÅstveit]

i try to do an object based supervides image classification and i have some problems. 

1. Should i make training data from the image, or from the segmentet raster?

2. I have made training data on sample mananger and done the Analysis Object based, random trees. On the same image as i made the trening samples this work fine, and the classification goes true. But i want to classify anotther image, is this possible? and how? because i only get failed message, that the input raster do not have any features. 

Helena 

[DanPatterson]

Are you following the recommendations cited in this link?

Understanding segmentation and classification—ArcGIS Pro | Documentation

[HelenaÅstveit]

yes, but i want to make training samples on one image, and then i want to classify another pitchure. in that way a person who know the area could make traning samples in one image, and then someone else can use that treinieng samples to klassify another pitchure where they dont know. 

[DanielMiranda2]

Deep Learning is probably a better approach for what you are trying to do:

An Overview of the Deep Learning Toolset 

Using the standard image classification tools, training samples have to be based on the image you want to classify.

[HelenaÅstveit]

Okei, Thank you!