Table 4 Accuracy of aggregate classifications of species in 4,149 image sets in comparison with expert classifications

From: Snapshot Serengeti, high-frequency annotated camera trap images of 40 mammalian species in an African savanna

Species

* Correct

Total

Proportion Correct

wildebeest

1,727

1,740

0.993

zebra

885

888

0.997

buffalo

244

252

0.968

hartebeest

245

251

0.976

Thomson’s gazelle

197

199

0.990

impala

172

181

0.950

other bird

101

134

0.754

warthog

120

120

1.000

giraffe

92

92

1.000

elephant

84

85

0.988

human

72

75

0.960

Grant’s gazelle

55

67

0.821

guinea fowl

55

56

0.982

spotted hyena

55

55

1.000

hippopotamus

28

28

1.000

eland

23

25

0.920

reedbuck

22

25

0.880

baboon

22

22

1.000

topi

15

20

0.750

lion (female&cubs)

18

18

1.000

dik dik

8

8

1.000

porcupine

8

8

1.000

cheetah

6

6

1.000

mongoose

5

6

0.833

serval

6

6

1.000

aardvark

4

4

1.000

bushbuck

3

4

0.750

kori bustard

4

4

1.000

secretary bird

4

4

1.000

jackal

1

3

0.333

leopard

3

3

1.000

ostrich

3

3

1.000

vervet monkey

3

3

1.000

aardwolf

1

2

0.500

hare

1

1

1.000

lion (male)

1

1

1.000

rhinoceros

1

1

1.000

rodent

0

1

0.000

waterbuck

1

1

1.000

  1. Total indicates the number of image sets containing that species, as identified by experts
  2. *Correct indicates the subset of those image sets for which the plurality algorithm arrived at the correct answer. Proportion Correct is given by # Correct / Total. Note that the 30 image sets marked as ‘impossible’ by experts do not appear in this table. Twenty-nine image sets were marked impossible due to insufficient amount of detail to make a definitive identification; one image set was marked impossible because it contained an animal (duiker) not in the list of 48 species available to the volunteers.