4.5 Merging Image Composite Sources

As stated in the beginning of the chapter, a composite source can be created from closed composite sources. For creating a composite from any existing sources, please see section 4.1 . A good application of this is to merge image composite sources. Here, “merge” refers to the concatenation of all the rows from each component source inside the created composite source.

For instance, images are labeled separately sometimes. One class of images were processed and uploaded to the BigML platform. Then another class of images were uploaded. They need to be merged to create datasets for machine learning.

When creating a composite source, the component sources should have the same fields, in other words they are homogeneous, so that the composite source can inherit all fields, the same as each component source. This maintains the “format” of the composite source (section 2.2 ).

If the component sources don’t have the same fields, or they are heterogeneous, the resulting composite source will have “Mixed” format, which cannot be used to create datasets.

Therefore, when merging image composite sources, it is very important to make sure that they have identical fields.

To create a composite source from existing composite source, use the Create composite source button on the action bar in the source list view as shown below.

Next, select the component sources by using the checkboxes next to them. In this case, select the existing image composite sources for the purpose of merging them.

Then click on the Create composite source button to create the new composite source.