[question] upscaling vs. parameter change for small target detection

[huemerj]

The target detection works quite well if a target is big enough, compared to the size of the whole image. If this is not the case, i.e. if the target is really small (30px in a 1400px image), the target is not detected anymore. So in order to detect a target in a comparatively big image, I can crop the target and then up-scale the cropped image part. Then I can detect the target reliably in the up-scaled image part.

My question is now: Is there are more feasible and natural way to detect small targets? By e.g. changing the default parameters? If yes, which parameters are responsible for multi-scale detection? Thank you for your suggestions!

[MrIsland]

Hi !

Did you ever manage to solve this question ? I happen to have the same problem as you and want to ask if anyone else knows of the solution to this problem?

Sorry to bother you, many thanks.

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.