text
stringlengths
0
820
of the LR and HR datasets. In both cases, 3 model runs have
been trained and we report both the average and the standard
deviation of our results to limit the variability coming from
the stochastic nature of the optimization.
The numerical evaluation is achieved using the accuracy,
defined in our case as the ratio of correct answers. We report
the accuracy per question type (see subsection II-A), the
average of these accuracies (AA) and the overall accuracy
(OA).
We show some predictions of the model on the different
test sets in Figure 8 and Figure 9 to qualitatively assess the
results. Numerical performance of the proposed model on the
LR dataset is reported in Table II and the confusion matrix
is shown in Figure 10. The performance on both tests sets of
the HR dataset are reported in Table III and the confusion
matrices are shown in Figure 11.
General accuracy assessment:
The proposed model achieves an overall accuracy of 79%
PRE-PRINT. FINAL VERSION IN IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 8
What is the area covered by residential buildings?
0 m² 0 m²Ground truth Prediction
(a) HR, test set 1
What is the area covered by rectangular buildings?
Between 100 and
1000 m²Between 100 and
1000 m²Ground truth Prediction
(b) HR, test set 1
Is there a residential building at the bottom
of the place of worship?
yes yesGround truth Prediction
(c) HR, test set 1
How many residential buildings at the bottom
of a road are there?
4 3Ground truth Prediction
(d) HR, test set 1
What is the amount of large buildings?
3 1Ground truth Prediction
(f) HR, test set 1
How many buildings on the left of a road
are there in the image?
38 0Ground truth Prediction
(e) HR, test set 1
What is the area covered by rectangular parkings?
between 100m2 and
1000m2between 100m2 and
1000m2Ground truth Prediction
(g) HR, test set 2
What is the amount of small buildings?
9 0Ground truth Prediction
(h) HR, test set 2
What is the amount of large residential buildings?
2 1Ground truth Prediction
(i) HR, test set 2
Fig. 8. Samples from the high resolution test sets: (a)-(f) are from the first set of the HR dataset, (g)-(i) are from the second set of the HR dataset.
PRE-PRINT. FINAL VERSION IN IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 9
TABLE II
RESULTS ON THE TEST SET OF THE LOW RESOLUTION DATASET . THE
STANDARD DEVIATION IS REPORTED IN BRACKETS .
Type Accuracy
Count 67.01% (0.59%)
Presence 87.46% (0.06%)
Comparison 81.50% (0.03%)
Rural/Urban 90.00% (1.41%)
AA 81.49% (0.49%)
OA 79.08% (0.20%)
TABLE III
RESULTS ON BOTH TEST SETS OF THE HIGH RESOLUTION DATASET . THE
STANDARD DEVIATION IS REPORTED IN BRACKETS .
Type Accuracy Accuracy
Test set 1 Test set 2
Count 68.63% (0.11%) 61.47% (0.08%)
Presence 90.43% (0.04%) 86.26% (0.47%)
Comparison 88.19% (0.08%) 85.94% (0.12%)
Area 85.24% (0.05%) 76.33% (0.50%)
AA 83.12% (0.03%) 77.50% (0.29%)
OA 83.23% (0.02%) 78.23% (0.25%)
on the low resolution dataset (see Table II) and of 83% on
the first test set of the high resolution dataset (Table III),
indicating that the task of automatically answering question
based on remote sensing images is possible. When looking at
the accuracies per question type (in Tables II and III), it can
be noted that the model performs inconsistently with respect
to the task the question is tackling: while a question about the
presence of an object is generally well answered (87.46% in
the LR dataset, 90.43% in the first test set of the HR dataset),
counting questions gives poorer performances (67.01% and
68.63% respectively). This can be explained by the fact that
presence questions can be seen as simplified counting ques-
tions to which the answers are restricted to two options: ”0”
or ”1 or more”. Classical VQA models are known to struggle
with the counting task [38]. An issue which partly explains
these performances in the counting task is the separation of
connected instances. This problem has been raised for the case
of buildings in [33] and is illustrated in Figure 8(f), where the
ground truth is indicating three buildings, which could also be
only one. We found another illustration of this phenomenon
in the second test set in Figure 8(i). This issue mostly arises
when counting roads or buildings.
Thanks to the answers’ quantization, questions regarding the
areas of objects are generally well answered with an accuracy