st2berry
Reputation: 33
ESRGAN Training Input
I want to use ESRGAN and train it to upscale satellite images. These images are:
- 16m/pixel (resampled)
- RGB only
- Sentinel-2 L2A
- Is it worth providing a training dataset composed of .png files or .tiff?
- Also, I tried to download some satellite images in .tiff format, and they appear way too dark. When I open them through SNAP, they look over-saturated. How will that affect training? For reference, I have provided some images below.
Opened through Windows Photo Viewer
Opened through SNAP
How will that affect training? Any help is truly appreciated.
Upvotes: 1
Views: 28