Table 4 Detection results while training two versions of FIN-DETECT with respect to HADD and EADD.

From: FIN-PRINT a fully-automated multi-stage deep-learning-based framework for the individual recognition of killer whales

Metric

Dataset

HADDa

EADDb

Validation [%]

Test [%]

Validation [%]

Test [%]

Recall

95.0

89.0

95.1

94.4

Precision

80.0

85.0

94.2

94.1

F1-Score

86.9

87.0

94.7

94.2

mAPc

92.0

82.0

94.2

93.4

  1. aHADD Human-Annotated Detection Dataset.
  2. bEADD Extended-Annotated Detection Dataset.
  3. cmAP mean Average Precision.
Back to article page