Previewing predictions

Below you can find ways of previewing predictions generated by the various frameworks used through this tutorial. For most of them, you can use ADAMS' Preview browser and a specific content handler. Configurations for these handlers can be applied as follows:

Copy a configuration from below to the clipboard.
Open the Preview browser in ADAMS (under the Visualization menu).
Click on the ... button at the bottom to bring up the options dialog.
Click on the ⋁ button at the top right of the dialog and select Paste setup.
Accept the setup by clicking on the OK button.

Domains#

Image classification#

CSV#

adams.gui.tools.previewbrowser.ImageClassificationHandler -image-reader adams.data.io.input.JAIImageReader -reader "adams.data.io.input.ImageClassificationSpreadSheetReportReader -input ${HOME}/temp/preview/image_01965.csv -reader \"adams.data.io.input.CsvSpreadSheetReader -data-row-type adams.data.spreadsheet.DenseDataRow -spreadsheet-type adams.data.spreadsheet.DefaultSpreadSheet\"" -format "$ ({score|.2})"

Example:

Preview browser: Image classification

JSON#

The following content handler overlays the label and score on the image:

adams.gui.tools.previewbrowser.ImageClassificationHandler -image-reader adams.data.io.input.JAIImageReader -reader adams.data.io.input.ImageClassificationJsonReportReader -format "$ ({score|.2})"

Example:

Preview browser: Image classification

Object detection#

ROIs CSV format#

adams.gui.tools.previewbrowser.ObjectAnnotationsHandler -image-reader adams.data.io.input.JAIImageReader -file-suffix -rois.csv -reader "adams.data.io.input.ObjectLocationsSpreadSheetReader -input ${HOME}/temp/preview/A11_jpg.rf.ff2610c21c7f6d0a793cb58efc3bd96e-rois.csv -reader \"adams.data.io.input.CsvSpreadSheetReader -data-row-type adams.data.spreadsheet.DenseDataRow -spreadsheet-type adams.data.spreadsheet.DefaultSpreadSheet\" -row-finder adams.data.spreadsheet.rowfinder.AllFinder -col-left x0 -col-top y0 -col-right x1 -col-bottom y1 -col-polygon-x poly_x -col-polygon-y poly_y -col-type label_str -range-meta-data score" -cleaner "adams.gui.visualization.object.objectannotations.cleaning.OverlapRemoval -finder adams.data.objectfinder.AllFinder -overlap-detection adams.data.objectoverlap.AreaRatio -overlap-removal \"adams.data.overlappingobjectremoval.KeepHighestMetaDataValue -score-key score\"" -shape-plotter adams.gui.visualization.object.objectannotations.shape.NoShape -shape-color "adams.gui.visualization.object.objectannotations.colors.PerType -color-provider adams.gui.visualization.core.DefaultColorProvider" -outline-plotter adams.gui.visualization.object.objectannotations.outline.PolygonOutline -outline-color "adams.gui.visualization.object.objectannotations.colors.PerType -color-provider adams.gui.visualization.core.DefaultColorProvider" -label-plotter "adams.gui.visualization.object.objectannotations.label.Anchored -format \"$ ({score|.2})\"" -label-color "adams.gui.visualization.object.objectannotations.colors.PerType -color-provider adams.gui.visualization.core.DefaultColorProvider" -show-object-panel true

Example:

Preview browser: Object detection (ROIs CSV)

OPEX format#

adams.gui.tools.previewbrowser.ObjectAnnotationsHandler -image-reader adams.data.io.input.JAIImageReader -file-suffix .json -reader "adams.data.io.input.OpexObjectLocationsReader -input ${HOME}/temp/preview/A11_jpg.rf.ff2610c21c7f6d0a793cb58efc3bd96e.json" -cleaner "adams.gui.visualization.object.objectannotations.cleaning.OverlapRemoval -finder adams.data.objectfinder.AllFinder -overlap-detection adams.data.objectoverlap.AreaRatio -overlap-removal \"adams.data.overlappingobjectremoval.KeepHighestMetaDataValue -score-key score\"" -shape-plotter adams.gui.visualization.object.objectannotations.shape.NoShape -shape-color "adams.gui.visualization.object.objectannotations.colors.PerType -color-provider adams.gui.visualization.core.DefaultColorProvider" -outline-plotter adams.gui.visualization.object.objectannotations.outline.PolygonOutline -outline-color "adams.gui.visualization.object.objectannotations.colors.PerType -color-provider adams.gui.visualization.core.DefaultColorProvider" -label-plotter "adams.gui.visualization.object.objectannotations.label.Anchored -format \"$ ({score|.2})\"" -label-color "adams.gui.visualization.object.objectannotations.colors.PerType -color-provider adams.gui.visualization.core.DefaultColorProvider" -show-object-panel true

Example:

Preview browser: Object detection (OPEX)

Instance segmentation#

ROIs CSV format#

adams.gui.tools.previewbrowser.ObjectAnnotationsHandler -image-reader adams.data.io.input.JAIImageReader -file-suffix -rois.csv -reader "adams.data.io.input.ObjectLocationsSpreadSheetReader -input ${HOME}/temp/preview/Abyssinian_10-rois.csv -reader \"adams.data.io.input.CsvSpreadSheetReader -data-row-type adams.data.spreadsheet.DenseDataRow -spreadsheet-type adams.data.spreadsheet.DefaultSpreadSheet\" -row-finder adams.data.spreadsheet.rowfinder.AllFinder -col-left x0 -col-top y0 -col-right x1 -col-bottom y1 -col-polygon-x poly_x -col-polygon-y poly_y -col-type label_str -range-meta-data score" -cleaner "adams.gui.visualization.object.objectannotations.cleaning.OverlapRemoval -finder adams.data.objectfinder.AllFinder -overlap-detection adams.data.objectoverlap.AreaRatio -overlap-removal \"adams.data.overlappingobjectremoval.KeepHighestMetaDataValue -score-key score\"" -shape-plotter "adams.gui.visualization.object.objectannotations.shape.FilledPolygon -bounding-box-fallback-ratio 0.1" -shape-color "adams.gui.visualization.object.objectannotations.colors.PerType -color-provider \"adams.gui.visualization.core.TranslucentColorProvider -alpha 64 -provider adams.gui.visualization.core.DefaultColorProvider\"" -outline-plotter adams.gui.visualization.object.objectannotations.outline.PolygonOutline -outline-color "adams.gui.visualization.object.objectannotations.colors.PerType -color-provider adams.gui.visualization.core.DefaultColorProvider" -label-plotter "adams.gui.visualization.object.objectannotations.label.Anchored -format \"$ ({score|.2})\"" -label-color "adams.gui.visualization.object.objectannotations.colors.PerType -color-provider adams.gui.visualization.core.DefaultColorProvider" -show-object-panel true

Example:

Preview browser: Instance segmentation (ROIs CSV)

Image segmentation#

Indexed PNG format#

The following handler works for indexed PNGs and for grayscale ones:

adams.gui.tools.previewbrowser.SimpleImageSegmentationAnnotationsHandler -reader adams.data.io.input.JAIImageReader -overlay-reader "adams.data.io.input.PNGImageReader -color-provider adams.gui.visualization.core.DefaultColorProvider" -colorizer "adams.data.image.transformer.GrayOrIndexedColorizer -color-provider adams.gui.visualization.core.DefaultColorProvider"

Example:

Preview browser: Image segmentation (indexed PNG)

Generic#

Plain text#

Any text file can be viewed with this handler:

adams.gui.tools.previewbrowser.PlainTextHandler

JSON#

This handler displays JSON in a pretty printed way:

adams.gui.tools.previewbrowser.JsonPrettyPrintHandler

Example:

Preview browser: JSON