Advanced detection settings for tracking multiple unmarked subjects

Aim

To set EthoVision to track multiple subjects per arena, when the subjects are not color-marked.

▪To minimize subject identity swaps, make sure that the subjects are well detected in the whole arena. Make lighting as uniform as possible, and use indirect light, to minimize shadows.

▪When subjects crowd closely together EthoVision may temporarily assign the same position to those subjects, or lose track of one of them. This can be minimized or totally eliminated by setting good detection settings. Make sure that the body of the subjects is fully detected, and that the reference image does not contain the image of the subjects.

This topic applies to:

Experiments set to:

▪Number of Subjects per Arena: 2 or more.

▪Tracked Features: Center-point, nose-point and tail-base detection.

▪Body point detection technique: Contour-based

If you use Deep learning to track two subjects, you do not need to adjust detection settings. Simply choose the sample rate (step 1 below) and skip the remaining steps. See Deep learning: Requirements

Procedure

1.In the Detection Settings window, under Video, select the image source and sample rate. See Video file, image quality and sample rate.

 

2.Use the Automated Setup. When the Automated Setup gives good detection, proceed with the next step. If not, create advanced detection settings, follow the procedure in Advanced detection settings: Method.

 

3.In the Detection Settings pane, under Subject Identification, select Unmarked subjects.

 

4.optional  Click a cell under Body fill and specify the color for a subject. This helps checking identity swaps during tracking.

5.Adjust Advanced detection settings: Smoothing settings.

 

6.When tracking the three body points: adjust Subject contour for nose-tail base detection. In all other cases see Advanced detection settings: Subject contour.

 

7.Adjust Advanced detection settings: Subject size (multiple animals per arena).

 

Notes

▪When tracking unmarked animals, it is not possible to specify which animal is labeled “Subject 1”, which “Subject 2” and so on. EthoVision XT chooses which individual is assigned a label. This also means that the same animal in video 1 is not necessarily labeled with the same name in video 2. To link the identity (or role) of Subject 1, Subject 2, etc. enter that information as an independent variable in the Trial List. See Define an independent variable

▪When tracking unmarked subjects, identity swaps may occur, that is, what is labeled Subject 1 may at some point in the track be labeled as Subject 2, and what was Subject 2 is now Subject 1. Always check carefully the tracks, and, if necessary, correct the identity swaps. See Swap subjects