Advanced detection settings: Subject size (multiple animals per arena)
Aim
To specify the average (and variation) of size of subjects, when tracking multiple subjects per arena.
important When setting the Subject size, make sure the animals do not touch each other.
To set subject size
1.In the Subject size section, click the Advanced button.
In the Size window, in the figure at the top, the thin red contour represents the current size of what EthoVision XT assumes is the animal shape.
▪To set the Detected subject size, proceed with step 2. Set the Detected subject size to ensure that small or very large objects are not detected as a subject.
▪To set the Modeled subject size, proceed with step 3. Set the Modeled subject size to obtain an estimate of the subject’s body size, and body position of the different subjects when they overlap or touch.
2.Set the Minimum and Maximum subject size (represented by a green contour):
▪Maximum subject size: The largest surface area (in pixels) that is detected as the subject. Objects bigger than the Maximum subject size, for example, the experimenter's arm, are detected as noise and not tracked. Decrease the Maximum subject size until its thick green contour surrounds the thin red contour by a fair margin.
▪Minimum subject size: The smallest surface area (in pixels) that is detected as the subject. Objects smaller than the Minimum subject size, such as droppings or disturbed sawdust, are detected as noise and not tracked. Increase the Minimum subject size until its thick green contour is smaller than the thin red contour by a fair margin.
The two sliders are interdependent. So, after you have set the Minimum subject size, when you next change the Maximum subject size, the slider for the Minimum subject size also moves (although the size in pixels stays the same).
3.In the Modeled subject size group, select Apply settings to all subjects if your multiple animals have similar sizes.
4.Select one of the subjects to model the subject size for, by clicking the name of the subject.
5.Next, adjust the modeled subject size (under Average - pixels) to the detected subject size (under Current - pixels):
You do this by clicking the Grab button. Keep clicking the Grab button until the modeled (Average) subject size equals the detected (Current) subject size.
Below: Part of the Modeled subject size group in the Size window (left) and the Video window. In the table, Current shows the current detected subject size in pixels, Average shows the modeled subject size in pixels. The arrows point to the visual feedback you get about the current and average subject size in the Video window.
When the modeled (Average) subject size equals the detected (Current) subject size, the following becomes visible:
▪In the Modeled subject size group: the Average subject size now equals or is larger than the Current subject size (see the table in the figure below).
▪In the Video window: the modeled subject size now completely overlaps with the current subject size (see the figure below).
▪In the Size window, the bold yellow contour represents the modeled subject size. This now coincides with the detected subject size indicated by the thin red contour.
Below: Part of the Modeled subject size group in the Size window (left) and the Video window. The modeled (Average) subject size is now adjusted to the detected (Current) subject size. Compare the table and video window in this figure with those in the figure above.
6.You can now set the Tolerance. Click the corresponding cell and enter a value.
The Tolerance determines the deviation of the average subject size. When the Current detected size deviates more from the Average subject size than the Tolerance, then the object is not considered to be the subject anymore and EthoVision starts making an educated statistical guess about the body contour of the animal.
This is visible in the Video window by a wobbling marker-color area. When this happens when animals do not touch, you should increase the Tolerance.
7.Select the Fix check box for each subject.
8.Optionally click Advanced Modeled Parameters and change the settings for the Maximum noise size, Shape stability, and Modelling effort (see below).
Subject size when tracking multiple fish
Click Grab until the body fill color covers the whole body. Or, enter the body size under Average and select Fix. Keep Tolerance high (70-100%).
Make sure that in the Subject size window, the red contour oscillates within the limits of the Average subject size (bold yellow contour).
If that is not the case, click Grab when you see in the Video window that the subjects are detected correctly, or enter the appropriate value of size in the Average cell, then select Fix.
Advanced Modeled Parameters
Maximum noise size
Set the Maximum noise size. The value should be lower than the minimum subject size and but high enough to make sure that small objects with the same contrast as the subject are not detected.
The Shape stability setting is used when you track animals whose body can be occluded by, for example, cage bars or part of the body of another animal. When this happens, the animal's body consists of two separate objects that are close together.
The Shape stability optimized for slider has two extreme settings:
▪Occlusions: When you set the slider close to Occlusions, EthoVision considers separate objects that are close together part of one animal.
▪Noise: When you set the slider close to Noise, EthoVision considers separate smaller parts not part of the animal.
The figure below shows the animal model as a result of applying the two extreme Shape stability settings. 'Noise' shows that the front of the animal, on the other side of the bar, is not considered to be part of the animal. 'Occlusion' displays the animal body as a whole.
If you are not sure which setting to select, leave Shape stability at the default value of 620.
The Modeling effort setting is used when two animals touch and EthoVision loses the separate shapes. At this point, EthoVision tries to determine which part of the big 'merged' body fill belongs to either animal. This costs a lot of processing load.
The Modeling effort optimized for slider has two extreme settings:
▪Performance: When you set the slider close to Performance, EthoVision is only allowed a short time to determine which part of the 'merged' body fill belongs to which animal. Therefore, Modeling quality is low.
▪Modelling: When you set the slider to Modelling, EthoVision XT is allowed a longer time per frame to determine which part of the 'merged' body fill belongs to which animal. Therefore, Modelling quality is good, but this costs a lot of processor load.
We recommend to select Modelling only when you have a computer that exceeds the minimum system requirements.
When you are not sure which setting to select, leave Modeling effort at the default value of 500.
Tips for setting the Subject Size
▪Make sure you do not set the Tolerance too small; it is better get a wrong body size/shape than a wrong location of the animal.
▪It is better to set the Average subject size slightly bigger than the actual subject size, especially when you carry out nose-tail tracking.
▪If you want to carry out live tracking with multiple similarly-sized animals, it is recommended to first introduce one animal into the arena and make the Subject Size settings for this animal.
▪If the subject size changes a lot between trials, it is recommended to create new Detection Settings for this new size.