Table 9 STBRNN performance under adversarial conditions.

From: Bio-inspired motion detection models for improved UAV and bird differentiation: a novel deep learning framework

Condition

Precision

Recall

Accuracy

False positives (%)

Normal (baseline)

96.4%

94.8%

96.2%

3.8%

Low-light conditions

94.1%

91.7%

93.5%

5.2%

Occlusions

91.5%

88.9%

90.3%

7.1%

Sensor noise (Gaussian)

89.8%

87.2%

88.7%

8.4%