Closed emmajreid closed 1 week ago
Not sure where you found more LR bands. Maybe it makes sense to "make" them after having super-resolved the standard 6 LR bands? However, technically speaking you can feed 6 LR bands to the network at a time (with duplications as well) to get some kinf of output, not sure how useful that would be.
I'm attempting to apply it to non-S2 imagery :). But I'll try out passing them in 6 at a time (or perhaps in fewer batches, for science), and let you know how it goes.
How would you recommend applying the pretrained DSen2 model to images with more than 6 LR bands? I've tried just forcing it through, but get a shape incompatibility for obvious reasons.