Skip to main content

Table 7 Residential scene classification accuracy across studies using deep CNNs transfer learning models

From: Residential scene classification for gridded population sampling in developing countries using deep convolutional neural networks on satellite imagery

References

Scene class

Dataset

# Classes

Accuracy (%)

Relative scene accuracy ranking

Hu et al. [43]

Sparse Residential

UC Merced

21

85

19 of 21*

Med. Residential

UC Merced

21

85

19 of 21*

Dense Residential

UC Merced

21

90

17 of 21**

Han et al. [42]

Sparse Residential

UC Merced

21

95

12 of 21***

Med. Residential

UC Merced

21

90

19 of 21****

Dense Residential

UC Merced

21

85

21 of 21

Residential

SIRI-WHU

12

93

10 of 12*****

Residential

WHU-RS

19

88

19 of 19

Chew et al. (in this study)

Residential

Nigeria

2

94.5

NA

Residential

Guatemala

2

96.4

NA

  1. *Medium and Sparse residential tied for 19th/20th place
  2. **Tied with “intersection” for 17th/18th place
  3. ***Tied with seven other classes for 12th–18th place
  4. ****Tied with “storage tank” for 19th/20th place
  5. *****Tied with “idle land” for 10th/11th place