Compare two segmentations with Hoover metrics


This application compares a machine segmentation (MS) with a partial ground truth segmentation (GT). The Hoover metrics are used to estimate scores for correct detection, over-segmentation, under-segmentation and missed detection.

The application can output the overall Hoover scores along with coloredimages of the MS and GT segmentation showing the state of each region (correct detection, over-segmentation, under-segmentation, missed).

The legend for the colored images is as follow:

  • white: background

  • green: correct detection

  • magenta: over-segmentation

  • cyan: under-segmentation

  • red: missed detection (only for GT)

The Hoover metrics are described in: Hoover et al., “An experimental comparison of range image segmentation algorithms”, IEEE PAMI vol. 18, no. 7, July 1996.

This application has several output images and supports “multi-writing”. Instead of computing and writing each image independently, the streamed image blocks are written in a synchronous way for each output. The output images will be computed strip by strip, using the available RAM to compute the strip size, and a user defined streaming mode can be specified using the streaming extended filenames (type, mode and value). Note that multi-writing can be disabled using the multi-write extended filename option: &multiwrite=false, in this case the output images will be written one by one. Note that multi-writing is not supported for MPI writers.


Input ground truth -ingt image Mandatory
A partial ground truth segmentation image.

Input machine segmentation -inms image Mandatory
A machine segmentation image.

Background label -bg int Default value: 0
Label value of the background in the input segmentations

Overlapping threshold -th float Default value: 0.75
Overlapping threshold used to find Hoover instances.

Colored ground truth output -outgt image [dtype]
The colored ground truth output image.

Colored machine segmentation output -outms image [dtype]
The colored machine segmentation output image.

Correct detection score -rc float Mandatory
Overall score for correct detection (RC)

Over-segmentation score -rf float Mandatory
Overall score for over segmentation (RF)

Under-segmentation score -ra float Mandatory
Overall score for under segmentation (RA)

Missed detection score -rm float Mandatory
Overall score for missed detection (RM)


From the command-line:

otbcli_HooverCompareSegmentation -ingt maur_GT.tif -inms maur_labelled.tif -outgt maur_colored_GT.tif uint8

From Python:

import otbApplication

app = otbApplication.Registry.CreateApplication("HooverCompareSegmentation")

app.SetParameterString("ingt", "maur_GT.tif")
app.SetParameterString("inms", "maur_labelled.tif")
app.SetParameterString("outgt", "maur_colored_GT.tif")
app.SetParameterOutputImagePixelType("outgt", 1)


See also

otbHooverMatrixFilter, otbHooverInstanceFilter, otbLabelMapToAttributeImageFilter