Help
Summary
General help
Change demo dataset
You can easily change the dataset you are working on by loading a demo one available in the 'Manage datasets > Load a demo dataset' menu.
Upload your datasets
You can also work with your own datasets, you just need to upload it through the 'Manage datasets > Upload your dataset' menu.
Then, you can upload a txtl or a jpgl file through the 'Manage datasets > Upload a link file' menu.
If you have loaded a jpgl file, you can also upload a pol file through the 'Manage datasets > Upload a area file' menu.
Note: see file formats for more informations.
File formats supported
-
The me format is used for datasets.
-
The jpgl format is used to associate an image to each item.
-
The txtl format is used to associate a description (textual) to each item. The separator is '-----separator-----'.
-
The pol format is used to describe areas in pictures.
Moving the 3D views
You can freely move 3D views using your mouse.
To rotate around the origin, drag the visualization with the left-click of your mouse.
To zoom in or out, hold down the SHIFT key and drag up and down your mouse with the left-click down.
Display of item information
When you select an item, by clicking on it on the Star coordinate view for example, some informations are displayed on the right side of the panel.
-
Basically, in the data tab, the item number is displayed, its class and its features value.
-
If a txtl file has been loaded, a text is displayed in the desription tab.
-
If a jpgl file has been loaded, an image is displayed in the preview tab.
Note#1: you can browse all the images using the Next and Previous buttons.
Note#2: if a pol file has also been loaded, areas will be displayed on the images.
Annotation
Add & remove classes
You can add and remove classes using the 'Class management' popup.
-
To remove a class, click on the 'Remove' button next to the class you want to delete.
-
To add a new class, enter the desired name and click on the 'Add' button.
Different ways of annotating items
You can annotate (add a label) to any item using the context menu of the right-click.
-
You can annotate on the Star coordinate view.
-
If a jpgl file has been loaded, you can annotate on the preview tab.
-
If a jpgl file and a pol file have been loaded, you can also select and annotate areas on the preview tab.
Export your annotations
You can export your annotations by using the 'Annotations > Export your annotations' menu.
A pop-up will appear with the list of all the annotations made so far so that you can easily copy them to your clipboard.
Auto-annotate unlabel items
If you have loaded the uncertainty view, and if you think the model is good enough, you can use the 'auto-annotate' function.
Each unlabel items will be labeled with the class the model thinks it belongs to.
The Star coordinate view
If you are not familiar with this representation, you should have a look at the concepts of the Star coordinate plot.
- Each little sphere stands for an item.
- The color of the items represents the class it belongs to. The grey ones are unnanotated items.
- Each component (feature) is representated by a green sphere and an axe.
Rotating & scaling component axis
On the 'Star coordinate' view, you can both rotate and scale the component axis.
To rotate an axis, just drag it using the left-click of your mouse.
To scale an axis (give it more weight), hold down the CTRL key and drag it using the left-click of your mouse.
Dimensionality reduction
The
Use the slider in the pop-up to choose the number of components you desire.
View optimization
The
The Uncertainty view
If you are not familiar with this representation, you should have a look at the concepts of the Uncertainty plot.
- Each big sphere represents a class with its unique color.
- Each little sphere stands for an unlabeled item.
- The position and the color of the unlabeled items are computed thanks to the predictions of the model.