After creating a split for a particular machine learning task with
given sets of data, the user is able to see the overview of the split.
There can be multiple splits for the same dataset and different tasks.
For example, a car dataset with 10000 images can have 8000 images for
training, 1000 for testing, and the rest 1000 for validation. It can
also have another split with varying numbers of training, testing, and
validation data.
Note that the number of "images" shown in the column might vary according to the type of experiment as well. For example, the number of labels will be considered for label class prediction.
To view the information of the split, just click on the three dots
present in the top right corner of each split and choose split info.
In the split information, the user can see the different images by filtering out different attributes.
For example, here I have filtered the to see the test data with the labels bear and dummies.
Then, I can see the following result:
The user can also view the split summary by clicking on "Summary" in split info.
For the label class prediction split, the user is able to view the
distribution of the different labels in the test, train, and validation
set.
Clicking any of the splits will give access to the creation of an experiment.
All the existing experiments can be viewed at the left sidebar. Here exp1 has been created.