Skip to main content
. 2020 Jul 2;20(13):3710. doi: 10.3390/s20133710

Figure 4.

Figure 4

Two examples of the manually annotated ellipses (a); the label images created for instance segmentation (b); the extracted ground truth ellipses (c) and the results (d). The filled part of the ellipses shows the identified orientation of the animals. Note the adjusted overlaps in (c), which allow a comparison with the predicted ellipses. The lower picture in (d) shows a faulty detection.