search

LulinZhang / SpS-NeRF

SparseSat-NeRF: Dense Depth Supervised Neural Radiance Fields for Sparse Satellite Images
BSD 3-Clause "New" or "Revised" License
25 stars 0 forks source link

DSM GENERATE #6

Open rihpeach opened 2 weeks ago

rihpeach commented 2 weeks ago

Hello and thank you for your excellent work. I'd like to ask why I need a DSM file of the data when configuring my own dataset, don't I need to generate this in the first place, and if I don't have the DSM of my own image area, can I still generate the DSM of this area?

LulinZhang commented 2 weeks ago

Hi, thank you for your interest in our work. SpS-NeRF requires a low-quality DSM to improve the result, as it uses very sparse input images, which is challenging for NeRF. If you don't have DSM as input, you can still generate DSM but the result would be disappointing. So it is strongly recommended to generate DSM following this: https://github.com/LulinZhang/SpS-NeRF?tab=readme-ov-file#12-generate-dense-depth

rihpeach commented 1 week ago

Thank you for your answer, can I use the dem of ASTER GDEM instead, but I feel that the resolution of the dem data above is a bit low?