You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Radiometric variability between (and even within) NAIP acquisitions due to sun angle and/or sensor calibration can be substantial. If different bands have similar biases, band ratios may help to normalize imagery and improve model generalization. Implementing this would just require adding a processing step after loading NAIP imagery and re-running the workflow with the new bands.
We may be able to evaluate this without even training a model just by sampling a large number of pseudo-invariant features across years and seeing how inter-year variance compares between raw bands and ratios.
The text was updated successfully, but these errors were encountered:
Radiometric variability between (and even within) NAIP acquisitions due to sun angle and/or sensor calibration can be substantial. If different bands have similar biases, band ratios may help to normalize imagery and improve model generalization. Implementing this would just require adding a processing step after loading NAIP imagery and re-running the workflow with the new bands.
We may be able to evaluate this without even training a model just by sampling a large number of pseudo-invariant features across years and seeing how inter-year variance compares between raw bands and ratios.
The text was updated successfully, but these errors were encountered: