Thanks to the authors for such a great dataset!
In the pre-training phase, ssl4eo has 13 channels, while bigearthnet, the dataset used for fine tuning, has 12 channels.
What we want to know is whether the authors retrain all PatchEmbed modules or preprocess the dataset so that they have the same number of channels
thanks again!
Thanks to the authors for such a great dataset! In the pre-training phase, ssl4eo has 13 channels, while bigearthnet, the dataset used for fine tuning, has 12 channels. What we want to know is whether the authors retrain all PatchEmbed modules or preprocess the dataset so that they have the same number of channels thanks again!