IEEE Geoscience and Remote Sensing Magazine - December 2023 - 85

4 September 2017 in Fort Bend County, TX, USA, as well as
other immediately affected locations. This dataset has been
extended in [119] with more images and questions.
TextRS-VQA
This dataset [113] is based on the TextRS [78] dataset proposed
for RS image captioning and retrieval. The dataset contains
2,144 images with varying spatial resolutions, and each image
is annotated with two to five question-answer pairs, for
a total of 6,245 pairs. Unlike the previous datasets, this dataset
is manually annotated with open-ended questions that
fall under four types: class name, presence, amount of an
object, and other. Therefore, this dataset is characterized by
having large diversity in the questions and answers.
TABLE 8. STATE-OF-THE-ART RESULTS OF DIFFERENT RS TEXT-IMAGE RETRIEVAL MODELS.
DATASET
UCM-captions
METHOD
DUM [79]
LW-MCR [81]
AMFMN [85]
SAM t-i [84]
KCR [89]
MAFA [92]
Wang et al. [82]
Rahhal et al. [87]
FBCLM [88]
CABIR [90]
RSICD
LW-MCR [81]
FCLM [83]
AMFMN [85]
KCR [89]
SAM t-i [84]
MAFA [92]
Rahhal et al. [87]
FBCLM [88]
CABIR [90]
GaLR [93]
EnVLF [95]
SWAN [96]
Sydney-captions
LW-MCR [81]
AMFMN [85]
SAM t-i [84]
MAFA [92]
RSITMD
Wang et al. [82]
FBCLM [88]
CABIR [90]
AMFMN [85]
LW-MCR [81]
NWPU-captions
TextRS
Rahhal et al. [87]
FBCLM [88]
GaLR [93]
EnVLF [95]
SWAN [96]
SAM t-i [84]
MAFA [92]
DUM [79]
DBTN [78]
R@: recall@; MEDR: median rank; mR: mean-recall.
DECEMBER 2023 IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE
85
R@1
21.86
13.14
12.86
13.8
17.43
10.3
17.11
19.33
27.33
12.67
4.3
9.11
4.9
5.4
11.5
12.9
9.14
13.54
5.42
4.69
6.09
5.56
15.52
14.83
9.6
13.1
25.65
27.1
19.66
9.96
9.25
17.61
10.44
11.15
11.15
11.24
14.8
13.5
19.02
12.55
R@5
60
TEXT TO IMAGE
R@10
50.38
53.24
55.2
57.52
48.2
62.19
64
72.67
54.19
18.85
31.61
18.28
22.44
35.7
32.4
28.96
38.74
20.77
19.48
22.33
22.26
58.28
56.55
34.6
61.4
61.71
70.32
57.59
34.03
34.07
49.73
37.01
36.68
38.63
40.4
45.2
39.4
55.25
41.62
75.58
79.52
79.43
94.8
80.38
80.1
93.03
91.42
94.38
89.23
32.34
50
31.44
37.36
53.4
47.6
44.59
56.94
33.58
32.13
36.65
37.41
80.34
77.89
55.8
81.9
81.41
89.68
78.51
52.96
54.03
66.59
57.94
51.68
57.58
60.4
82.9
67.1
71.72
62.09
MEDR
-
-
-
5
-
-
-
-
-
-
-
10
-
-
8
-
-
-
-
-
-
-
-
-
9
-
-
-
-
-
-
-
-
-
-
-
6
-
-
-
R@1
25.47
18.1
16.67
9.5
15.24
14.5
19.05
19.04
28.57
15.17
4.39
11.27
5.39
5.95
12.8
12.3
10.7
13.27
8.59
6.59
7.78
7.41
20.69
24.14
5.8
22.3
27.59
25.81
32.76
11.06
9.73
19.69
12.84
14.82
13.42
13.35
13.3
15.2
22.95
12.55
R@5
IMAGE TO TEXT
R@10
59.76
47.14
45.71
47.6
50.95
56.1
58.57
53.33
63.81
45.71
13.35
37.94
15.08
18.59
31.6
35.7
29.64
27.17
16.27
19.85
20.52
20.13
60.34
51.72
32.7
60.5
62.07
56.45
65.52
29.2
26.77
40.26
30.53
31.64
33.11
32.15
38.8
46
59.52
39.53
72.61
63.81
68.57
78.6
73.33
95.7
81.43
77.61
82.86
72.85
20.29
54.41
23.4
29.58
47.3
54.8
41.53
37.6
24.13
31.04
31.56
30.86
77.59
75.86
48.1
76.4
75.86
75.81
79.31
38.72
37.61
54.42
45.89
42.48
47.12
46.9
65.5
83.5
77.23
59.53
MEDR
-
-
-
6
-
-
-
-
-
-
-
8
-
-
9
-
-
-
-
-
-
-
-
-
11
-
-
-
-
-
-
-
-
-
-
-
7
-
-
-
mR
52.55
45.35
46.08
49.92
49.14
50.82
55.23
54.12
61.6
48.3
15.59
32.39
16.42
19.89
32.05
32.62
27.43
31.21
18.12
18.96
20.82
20.61
52.13
50.17
31.1
52.6
55.72
57.53
55.56
29.32
28.58
41.38
32.44
31.41
33.5
34.11
43.42
44.12
50.95
37.98
IEEE Geoscience and Remote Sensing Magazine - December 2023

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - December 2023
