Remote Sensing Change Detection via Weak Temporal Supervision
Authors: Xavier Bou, Elliot Vincent, Gabriele Facciolo, Rafael Grompone von Gioi, Jean-Michel Morel, Thibaud Ehret
Organizations: Université Paris-Saclay, CNRS, ENS Paris-Saclay, Centre Borelli, France · LASTIG, Université Gustave Eiffel, IGN-ENSG, France · Lingnan University, Hong Kong · AMIAD, Pole Recherche, France
Semantic change detection in remote sensing aims to identify land cover changes between bi-temporal image pairs. Progress in this area has been limited by the scarcity of annotated datasets, as pixel-level annotation is costly and time-consuming. To address this, recent methods leverage synthetic data or generate artificial change pairs, but transferring change detection models across datasets and acquisition conditions remains challenging. In this work, we introduce a weak temporal supervision strategy that leverages additional temporal observations of existing single-temporal datasets, without requiring any new annotations. Specifically, we extend single-date remote sensing datasets with new observations acquired at different times and train a change detection model by assuming that real bi-temporal pairs mostly contain no change, while pairing images from different locations to generate change examples. To handle the inherent noise in these weak labels, we employ an object-aware change map generation and an iterative refinement process. We validate our approach on extended versions of the FLAIR and IAILD aerial datasets, showing competitive zero-shot and low-data regime performance. Lastly, we showcase results over large areas in France, highlighting the scalability potential of our method.
Figures & tables
Figure 2: Visual examples. For each dataset, we show example triplets ( St , It , It′ ) from left to right: annotation mask, original image, and additional acquisition. Top row: significant land cover changes; middle and bottom: irrelevant changes (shadows, seasonal variations).
Data source
Model
Dataset extension
Bi-temporal
Post-classif.
Synthetic
STAR [ 80 ]
Ours
In-domain
Out-of-Domain
FLAIR-BCD
FLAIR-BCD
LEVIR-CD
WHU-CD
S2Looking
heldout
spot
F1
IoU
F1
IoU
F1
IoU
F1
IoU
F1
IoU
FLAIR
UNet
—
✗
✓
✗
✗
✗
53.1
36.1
—
—
51.4
34.6
66.7
50.1
10.6
5.6
UNet
b-FLAIR
✓
✓
✗
✗
✗
51.9
35.0
—
—
40.3
25.3
65.6
48.8
8.8
4.6
Dual UNet
Changen2-S19-180k [ 79 ]
✗
✗
✓
✗
✗
47.5
31.1
—
—
51.8
35.0
63.3
46.3
13.4
7.2
Table 1: Comparison with baselines. For each data source, the best-performing results are shown in bold . Second-best results are underlined . Note that SyntheWorld is a fully synthetic dataset, in contrast to FSC-180k or Changen2-S19-180k, which are built on FLAIR; therefore, evaluation on real data is always out-of-domain for models trained on SyntheWorld. † Results from [ 4 ] .
Figure 3: XOR vs. sIoU for change map generation from image pairs, for the building change detection binary task. Top: a real pair with slight viewpoint variation—XOR falsely detects changes, while sIoU correctly shows none. Bottom: a fake pair with different buildings—XOR misses overlapping changes, sIoU correctly marks both. Only one building per image is labeled for clarity.
Figure 4: Qualitative results on FLAIR-BCD. We compare the building change maps predicted by baseline methods and ours.
Pretraining
Proportion of target data used
dataset
1%
10%
30%
LEVIR-CD
No pre-training †
0.36
0.69
0.75
SyntheWorld †
0.41 (+14%)
0.71 (+3%)
0.77 (+3%)
FSC-180k †
0.55 (+53%)
0.73 (+6%)
0.79 (+5%)
Ours (b-FLAIR)
0.60 (+67%)
0.77 (+12%)
0.81 (+8%)
Table 2: Low data regimes. IoU scores on LEVIR-CD and S2Looking datasets when fine-tuning a Dual UNet on limited target data (1%, 10% and 30%). Each result is averaged over 10 runs. Bold denote the best result for each dataset. † Results from [ 4 ] .
Hi-UCD mini
Hi-UCD
Method
Ch. IoU
Seg. mIoU
Ch. IoU
Seg. mIoU
Post-classification + temp. aug. (UNet)
14.4
37.7
15.0
38.2
STAR [ 80 ] (Dual UNet)
45.1
26.3
45.3
48.3
Ours (Dual UNet)
47.2
39.7
47.2
46.7
Table 3: Results on the multi-class Hi-UCD mini dataset. Comparison between the single-temporal baseline, STAR and our method. Models are trained either on the Hi-UCD mini training set or on the full Hi-UCD training split.
Figure 5: Ablation on different backbones. F1 scores of our method compared to FSC-180k [ 4 ] and STAR [ 80 ] , for the DualUnet [ 4 ] and SCanNet [ 24 ] architectures.
Method
Focal loss
XOR
sIoU
Real pairs
F1↑
IoU↑
FPR↓
STAR [ 80 ]
✗
✓
✗
✗
66.3
49.6
2.51
✓
✓
✗
✗
67.4
50.8
2.42
✓
✗
✓
✗
72.6
56.6
1.37
✓
✓
✗
✓
76.0
61.3
0.60
Ours (1 Iter.)
✓
✗
✓
✓
77.0
62.6
0.45
Table 4: Comparison with STAR. Training on b-FLAIR subset and evaluating on FLAIR-BCD.
Table 5: Impact of preal and τ . Left: Comparison of different values for the proportion preal of real bi-temporal pairs. Right: Comparison between XOR change maps and our sIoU-based maps with different threshold τ . Scores reported on FLAIR-BCD.
FLAIR-BCD
FLAIR-BCD-spot
F1
IoU
F1
IoU
Iteration 1
67.3
50.7
22.6
12.7
Iteration 2
78.1
64.1
21.7
12.2
Iteration 3
79.0
65.2
22.9
12.9
Table 6: Performance across training iterations. F1 Score (%) and IoU (%) of a Dual UNet trained with our method over three training iterations, evaluated on the corresponding in-domain test set. Best results per dataset are shown in bold.
Figure 6: Samples removed during iterative refinement. Triplets ( It , It′ , M^ ) removed during refinement exhibit real building changes or blurred sensitive areas, where the assumption M=0 does not hold. Such samples are correctly identified by the model and excluded from the training set on subsequent iterations.
Method
%
# Im.
km 2
Ac t1
An t1
Ac t2
An t2
Gen
Ch. IoU
Seg. mIoU
Post-classif.
100
300
3.14
✓
✓
✗
✗
✗
14.4
37.7
STAR
100
300
3.14
✓
✓
✗
✗
✗
45.1
26.3
Changen2
90
270
2.92
✓
✓
✗
✗
✓
22.7
35.2
Ours
80
240
2.50
✓
✓
✓
✗
✗
47.2
40.0
Sup. CD
50
150
1.57
✓
✓
✓
✓
✗
44.0
34.9
Table A1 : Evaluation under an equal data budget on Hi-UCD mini. Reference budget is acquisition + annotation at t1 (i.e. Post-classif. and STAR). Synth adds cost of t2 image generation (Gen). Ours adds budget to acquire t2 , while supervised CD annotates t2 .
Figure B1 : Additional example triplets from our extended datasets , b-FLAIR, b-FLAIR-spot, and b-IAILD, to complement the visual examples already displayed in Fig. 2 of the main paper. For each dataset, we show triplets ( St , It , It′ ) in this order from left to right.
Figure B2 : Spatial repartition of FLAIR-BCD(-spot/-heldout) data w.r.t. FLAIR , within Metropolitan France. Note that even if pairs from FLAIR-BCD(-spot) are from departments already covered by FLAIR Train/Val, we made sure that there is no intersection between the two datasets. We also indicate with a ⋆⋆⋆⋆⋆⋆⋆⋆⋆⋆⋆⋆⋆⋆⋆⋆⋆ the location of the large-scale zone used for Fig. 1 of the main paper.
Figure B3 : Temporal difference for b-FLAIR pairs (left) and estimated variation of the change rate with temporal distance (right).
Figure B4 : Estimated variation of the change rate per AoIs/temporal distance on b-IAILD.
Figure B5 : Example triplets of our test sets for building change detection. We show additional example triplets ( It , It′ , M ) in this order from top to bottom, for FLAIR-BCD (a), FLAIR-BCD-spot (b), and FLAIR-BCD-heldout (for (c) same res. and (d) different res. pairs).
Dataset
Modality
Resolution (cm/px)
Location
Image Size
Num. test pairs
LEVIR-CD [ 13 ]
Aerial
50
Texas, US
1024×1024
128
WHU-CD [ 35 ]
Aerial
7.5
Christchurch, New Zealand
256×256
762
S2Looking [ 60 ]
Satellite
50-80
Global
1024×1024
1000
Hi-UCD mini [ 67 ]
Aerial
10
Estonia
1024×1024
386
FLAIR-BCD (ours)
Aerial
20
France
512×512
1730
FLAIR-BCD-spot (ours)
Satellite
160
France
64×64
1730
Table B1 : Overview of benchmark datasets. We report the main characteristics of the 6 evaluation datasets on which we compare our approach to the baselines.
Figure C1 : General diagram of the proposed methodology. Our methodology can be applied to any 3-branch SCD model, provided it can take two images at different dates as input and produce a semantic segmentation mask for each date, and a change mask. The semantic segmentation task is supervised through the single-temporal available ground truth, implying a temporal augmentation on the additional non-annotated image. The change detection task is supervised via weak temporal augmentation, creating fake change examples by pairing images of different locations, and no-change examples by assuming no changes occurred between real pairs.
Figure C2 : Example of generated change maps for pairs of images of b-FLAIR from different locations, with a focus on the “building class”. We show maps resulting from logical operations (XOR or OR) as well as sIoU-based maps computed for various values of the threshold τ . We highlight with red circles , areas for which differences in building footprints between the two images are variably considered as changed or not depending on the method.
Dataset
FID ↓
IS ↑
Changen-90k [ 81 ]
34.74 †
5.41 †
Changen2-S1-15k [ 79 ]
32.44 †
4.64 †
Changen2-S19-180k (Ours)
35.69
5.69
Table C1 : Image quality evaluation. We report image quality metrics for Changen2-based synthetic datasets (including ours). † Results reported in [ 79 ] .
Figure C3 : Example synthetic change pairs generated with our implementation of Changen2 on FLAIR. We show FLAIR images, their corresponding semantic mask, our generated change images, and the new semantic mask, in this order from top to bottom.
Method
F1↑
IoU↑
FPR↓
Post-classif
64.2
47.3
2.87
Post-classif + Temp. aug.
59.0
41.8
2.46
Changen2-S19-180k pretraining
58.5
41.4
3.63
FSC-180k pretraining
83.1
71.1
0.73
STAR
75.1
60.1
1.58
Ours
79.0
65.3
0.35
Table D1 : FLAIR-BCD results for methods trained on b-FLAIR. The FLAIR-BCD is introduced in Section B .
5x5
FPR ( ↓ )
Num.
Obj. size
median filter
obj. (avg)
(avg, m²)
Ground truth
—
—
1.7
187
FSC-180k pretraining [ 4 ]
✗
0.73
12.1
30
Ours (b-FLAIR)
✗
0.35
3.0
87
FSC-180k pretraining [ 4 ]
✓
0.71
3.6
98
Ours (b-FLAIR)
✓
0.28
1.8
145
Table D2 : Comparison between our method and a pretraining on FSC-180k [ 4 ] on FLAIR-BCD. We report the false positive rate (FPR) of both methods, as well as object-related indicators: average number of change objects per pair, and average size of change objects. An object is defined as a pixel-connected component in the binary building change map.
Statistic
Iter. 1
Iter. 2
Iter. 3
Total bitemporal pairs
61,712
61,228
60,680
Acquisition-gap distribution (%)
<24 months
2.2
2.3
2.3
24 – 30 months
7.0
7.0
7.0
30 – 36 months
35.1
35.1
35.1
36 – 42 months
47.3
47.2
47.2
Table D3 : Temporal statistics. Temporal statistics of remaining pairs before each training iteration.
Figure D3 : Geographic distribution of bitemporal pairs in b-FLAIR and filtering across refinement iterations.
Threshold (%)
Removed
Removed (%)
Acc (%)
1
1121
2.39
70.1
2
842
1.80
78.7
5
478
1.02
89.7
10
315
0.67
95.2
20
161
0.34
96.9
Table D4 : Filtering threshold. Removed samples for different pixel change thresholds for the third training iteration of the b-FLAIR model.
Dataset
F1
IoU
FLAIR-BCD
79.0
65.3
FLAIR-BCD-Inv
81.2
68.3
Table D5 : Results on FLAIR-BCD-inv. Inverted FLAIR-BCD input images to simulate building demolitions.
Figure D4 : Large scale qualitative comparison of FSC-180k and our method for an example of the Lille metropolitan area . Demolitions (blue) and constructions (orange) are shown in each case. As seen, FSC-180k predicts some false positives seemingly due to shadow changes and temporal variations.
Figure D5 : Large scale qualitative comparison of Changen2 and our method for the Lille metropolitan area . Demolitions (blue) and constructions (orange) are shown in each case. Changen2 seems to predict actual demolitions and construction events but it also generates an overwhelming number of false positives.
Data source
Model
Dataset extension
In-domain
Out-of-Domain
FLAIR-BCD-heldout
FLAIR-BCD-spot
LEVIR-CD
WHU-CD
S2Looking
F1↑
IoU↑
FPR↓
F1↑
IoU↑
FPR↓
F1↑
IoU↑
FPR↓
F1↑
IoU↑
FPR↓
F1↑
IoU↑
FPR↓
FLAIR
UNet
—
53.1
36.1
4.84
—
—
—
51.4
34.6
2.40
66.7
50.1
4.09
10.6
5.6
3.18
UNet
b-FLAIR
51.9
35.0
4.52
—
—
—
40.3
25.3
2.51
65.6
48.8
4.33
8.8
4.6
2.06
Dual UNet
Changen2-S19-180k [ 79 ]
47.5
31.1
5.89
—
—
—
51.8
35.0
4.20
63.3
46.3
4.46
13.4
7.2
2.11
Table D6 : Comparison with baselines. Similar to the table from the main paper but with the false positive rate reported for all methods. Best score for each training data source is highlighted in bold, and the second best score is underlined.
Data source
Iteration
Filtered samples
In-domain
Out-of-Domain
FLAIR-BCD
FLAIR-BCD-spot
LEVIR-CD
WHU-CD
S2Looking
F1↑
IoU↑
F1↑
IoU↑
F1↑
IoU↑
F1↑
IoU↑
F1↑
IoU↑
FLAIR
1
0
67.3
50.7
—
—
5.2 †
2.7 †
51.8
34.9
10.5
5.6
2
484
78.1
64.1
—
—
20.3 †
11.3 †
76.1
61.4
15.4
8.3
3
842
79.0
65.2
—
—
17.8 †
9.3 †
77.3
63.0
13.6
7.3
Table D7 : Performance across training iterations. F1 Score (%) and IoU (%) of our method over three training iterations. Best score for each dataset within a training data source is highlighted in bold. The number of filtered pairs with respect to the training dataset is indicated for each iteration. † Scores for LEVIR-CD with FLAIR data source where computed with a crop size of 256 without resampling (see Sec. D.7 ).
Figure D6 : Large-scale SPOT-6/7 change detection results. Comparison between ground truth, our b-FLAIR-spot model, and the STAR approach over the metropolitan region region of Lille, France. While our model predicts a considerable number of false negatives, it detects most areas where larger changes were produced, while the STAR version detects an overwhelming amount of false positives.
Figure D7 : Qualitative results on FLAIR-BCD-spot. We compare the building change maps predicted by baseline methods and ours.
Figure D8 : Qualitative zero-shot results on LEVIR-CD for 3 randomly selected input pairs. “PC” stands for “post-classification” and “p.t.” for “pre-training”.
Figure D9 : Qualitative zero-shot results on WHU-CD for 3 randomly selected input pairs. “PC” stands for “post-classification” and “p.t.” for “pre-training”.
Figure D10 : Qualitative zero-shot results on S2Looking for 3 randomly selected input pairs. “PC” stands for “post-classification” and “p.t.” for “pre-training”.
Crop size 256
Crop size 512
Resampling size
256
512
256
512
Post-classif.
37.3
51.4
10.4
37.4
Post-classif.+temp. aug.
32.3
40.3
7.9
30.4
Changen2-S19-180k
45.4
51.8
17.3
44.8
FSC-180k (our eval)
57.0
65.3
11.2
56.3
STAR
43.6
53.3
12.0
44.5
Table D8 : F1 scores on LEVIR-CD for different image sizes and resampling scales . As seen, simple interpolation to zoom into the images drastically increases performance.
Figure D11 : Qualitative examples of the Segment Any Change method with no prompt provided to indicate building semantics . As shown, class-agnostic changes are predicted all over images, yielding a considerable number of false positives. Top three rows include images from the main text Figure 1, and bottom examples correspond to FLAIR-BCD samples.
Figure D12 : Prompt-guided qualitative examples of the Segment Any Change method. A prompt, i.e. a localization point in one of the two input images, is given to indicate building semantics. As shown, prompt guiding does not provide competitive results. Top three rows include images from the main text Figure 1, and bottom examples correspond to FLAIR-BCD samples.
Figure D13 : Qualitative Hi-UCD results on binary change detection. We compare the change detection ground truth (CD-GT) to our approach, STAR and post-classification method with temporal augmentation (PC+TA). Green pixels indicate no change, red pixels indicate change, and black pixels are unlabeled. Complementary to quantitative results, our method seems to outperform baselines on binary change detection.
Figure D14 : Qualitative Hi-UCD results on semantic segmentation. We show segmentation predictions for timestamps t1 and t2 . Our approach is compared with STAR and the post-classification method with temporal augmentation (PC+TA). Our method proved slightly below STAR on semantic segmentation.