search

dongrunmin / RRSGAN

41 stars 4 forks source link

About test.py #6

Open StevenTTT123 opened 1 year ago

StevenTTT123 commented 1 year ago

my envs is run on autodl (a remote host) On that platform, installing slurm (to run 'srun' from test.py 8) seems really difficult. My question is, is there an easy way to run test.py directly since it's conducted on single computer (singe GPU) without installing and configuring slurm? Thank you so much.

StevenTTT123 commented 1 year ago

to make it clear, test.py refers to val.py (in RRSGAN) in your code.

dongrunmin commented 1 year ago

to make it clear, test.py refers to val.py (in RRSGAN) in your code.

In order to resolve your issue, please remove the following part from val.sh: "srun -p ${partition} --gres=gpu:1".

StevenTTT123 commented 1 year ago

Thank you so much. I have solved the problem by directly running val.py. Besides, I have an additional question. How did you collect your dataset? I mean , several different satellite imagesets (Google Earth for Ref, GF-X for HR ...) over the same scene? We are planning on collecting similar datasets over the same scenario (from satellite A, B, ...), by it seems like a huge work. Do you have any tips?

dongrunmin commented 1 year ago

Thank you so much. I have solved the problem by directly running val.py. Besides, I have an additional question. How did you collect your dataset? I mean , several different satellite imagesets (Google Earth for Ref, GF-X for HR ...) over the same scene? We are planning on collecting similar datasets over the same scenario (from satellite A, B, ...), by it seems like a huge work. Do you have any tips?

Hope the following steps can help you solve this problem. First, we can download images of the same cities from different platforms, such as Google Earth Engine and Microsoft Virtual Earth. Secend, these downloaded images come with geospatial coordinate information, which can be extracted using libraries like GDAL or RasterIO. Finally, we can crop both images to obtain a perfectly matching pair of images for the same region by utilizing the geospatial coordinates.

StevenTTT123 commented 1 year ago

Thank you so much. Wish u a nice day!