* How should we derive the RGB images from VV and VH images?
* We have segmentation maps for two things: water body level and flood level. For the sake of the competition, we are only supposed to be caring about the flood levels. Is there any mapping that VV images should be used for flood level prediction and VH images for water body level?
* Upon visualizing the data, we saw many segmentation maps (i.e. the ground-truths) are all black. Is this expected?
Hi, to answer your questions:
* How should we derive the RGB images from VV and VH images?
You can read more about the VV and VH bands here: https://sentinel.esa.int/web/sentinel/user-guides/sentinel-1-sar/acquisition-modes
These images are from different radar bands. You can learn more about their acquisition and conversion process here: https://nasa-impact.github.io/etci2021/
* We have segmentation maps for two things: water body level and flood level. For the sake of the competition, we are only supposed to be caring about the flood levels. Is there any mapping that VV images should be used for flood level prediction and VH images for water body level?
The water body labels are provided as an added information. It’s up to the participant if they want or don’t want to use them. Evaluation is done solely using flood predictions.
No, there is no such mapping b/w the bands and labels whatsoever. These images are from different radar bands and you can learn more about their acquisition and conversion process here: https://nasa-impact.github.io/etci2021/
* Upon visualizing the data, we saw many segmentation maps (i.e. the ground-truths) are all black. Is this expected?
Black segmentation maps mean that they are devoid of flood activity.
Let me know if you have other questions. Thanks!