- The checkpoints can be used to start the training in
train.py
(use the--load_ckpt
parameter). You can retrieve the training parameters values from the checkpoint file name. - The SavedModels can be used directly on remote sensing images from
sr.py
to generate high-resolution images. - Before applying a model, check that your input image has the same spectral content/bands + in the same order as indicated in the table below.
The model aims to upscale Sentinel-2 images from 10 meters to 2.5 meters, with the four spectral bands ordered in the following:
order | band |
---|---|
1 | red (B4) |
2 | green (B3) |
3 | blue (B2) |
4 | near infrared (B8) |
The model was trained from 250 different Spot-6 and Spot-7 scenes covering the entire France Mainland, acquired during the year 2020, from march to october, and the Sentinel-2 images acquired close to the same day. We used TOC Sentinel-2 products from the THEIA Land data center. Spot-6 and Spot-7 images were interpolated at 2.5 meters and radiometrically calibrated to match the Sentinel-2 radiometry. Around 150k patches were used to train the model.
You can download the pre-trained model here: