This work proposes a novel change detection data setting which uses both optical and SAR images pre-change, yet only SAR imagery post-change. For this challenging scenario, we propose a Siamese network that processes the pre-change and post-change SAR inputs using a shared set of weights, while the pre-change optical input is processed using a network that do not share the weights with the SAR inputs.
Before running the code, download the OSCD multi-sensor dataset as instructed in the paper https://www.int-arch-photogramm-remote-sens-spatial-inf-sci.net/XLIII-B3-2021/243/2021/isprs-archives-XLIII-B3-2021-243-2021.pdf (dataset link is provided in Section 3.1)
For training model: $ python -u mainAllChannelsSiameseComplex.py --manualSeed 19 (manusal seed can be changed)
Note that path to the OSCD multi-sensor dataset is set in line 48 and the path where the trained model will be stored is set in line 63. Please change them as appropriate.
After training, to compute statistics relevant for test-time adaptation,
$ python computeStatisticsFromTrainingImages.py
Set the path to the dataset in line 31 and path to where the folder where model is stored in 46
To evaluate scores, $ python evaluateScores.py
Set the path to the dataset in line 32
Set the statistics for test-time adaptation computed in previous step in line 42
Set where the model is stored in line 48