Knowledge Base

Problem

EthoVision fails to reliably detect an animal during tracking because the animal's fur color is too similar to the arena background, or because a tethered cable (e.g., fiber-optic or recording cable) disrupts detection.

Details

Affects all EthoVision XT versions. Common scenarios: (1) A black mouse on a dark arena floor with a black tether cable causes missed samples or the tether being tracked instead of the animal. (2) A white mouse on a light-colored arena surface is not detected at all, even with deep learning enabled. The deep learning/body-point module localizes tracking points on a detected subject but does NOT replace the underlying contrast-based detection—if the animal cannot be detected, deep learning will also fail.

Solution

**General principle:** EthoVision's primary detection relies on contrast between the animal and the background. Deep learning body-point placement is a secondary step and cannot detect an animal that the contrast algorithm misses.

**For poor contrast (animal color ≈ background color):**
1. Modify the arena/background to increase contrast:
- Use a darker floor insert or paper for white animals.
- Use a lighter floor insert for dark animals.
2. Adjust lighting: increase illumination intensity or angle to create shadow/edge contrast, or use IR lighting if the camera supports it.
3. In EthoVision Detection Settings, adjust the threshold and try 'Brightest' or 'Darkest' detection mode depending on whether the animal is lighter or darker than the background.
4. If multiple fur colors are present in the same experiment, create separate Detection Settings profiles for each color and apply them per trial or per arena.

**For tethered animals (cable interfering with detection):**
1. In Detection Settings, adjust the **minimum/maximum subject size** (pixel area) to exclude the thin tether cable while keeping the animal body within range.
2. Experiment with the **Smoothing** and **Erosion/Dilation** settings to suppress thin artifacts.
3. If the tether is unavoidable, consider using a lighter-colored cable sleeve to reduce its contrast with the background relative to the animal.

**Note:** If deep learning body-point tracking is still required, the animal must first be reliably detected by the standard contrast-based method; address detection before configuring body-point placement.