Thanks a lot for sharing your implementation and model weights for Sentinel2 cloud detection :)
I have a few questions concerning your implementation of the mobilenetv2 cloud detector.
in your L2A 4 bands version, it seems (here) that you use the "Red Edge 4" band (B8A) instead of the "NIR" band (B08). Is it correct? If so, why using B8A (sampled at 20m) instead of B08 (sampled at 10m like B02, B03, B04)?
Likewise, your 13 bands version uses 11 radiance bands (you removed B08 from the 12 L2A radiance bands) and you add the AOT and WVP bands. When used as input of you model, the first 11 bands should be divided by 10000 and the remaining 2 should be divided by 65535, correct? Again, is there a reason why you removed B08?
Hi,
Thanks a lot for sharing your implementation and model weights for Sentinel2 cloud detection :)
I have a few questions concerning your implementation of the mobilenetv2 cloud detector.
Thanks a lot for your help!
Best regards