DenseSTF

DenseSTF impelemented by TensorFlow 1.14

Attached file namely "data.mat" shows the high resolution landsat(hr1,hr2,hr3) and low resolution MODIS(lr1,lr2,lr3) images for cropland scene.

The pre-processing of cloudmask and gapfilling has been conducted in MATLAB. Surface reflectence for all images are ranging 0~0.6.

Orginal images in the "data.mat" have a size of 1000x1000. A subset of 500x500 was used for train and test, i.e, data[250:750,250:750,:]

Train:

python densestf_train_mband.py & python densestf_train_mband_lulc.py

Test:

python densestf_test_mband.py & python densestf_test_mband_lulc.py

Evaluation using MATLAB:

load('data.mat','hr2')

load('pred-densestf.mat','pred')

Metrics = computeMetric(hr2[251:750,251:750,:],pred); % matlab index starts from 1