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
Is your feature request related to a problem? Please describe.
A remote sensing focussed pre-trained model needs remote-sensing focused augmentations for the input imagery. A transform that mimics real-world noise would be one that simulates patchy clouds over images. This would have a similar effect to masking used in mainstream computer vision but more realistic.
Describe the solution you'd like
Clouds could be simulated by generating randomly placed patchy holes with Gaussians, then using a graduated alpha layer to simulate the Gaussian gaps in the clouds with smooth edges.
The text was updated successfully, but these errors were encountered:
Is your feature request related to a problem? Please describe.
A remote sensing focussed pre-trained model needs remote-sensing focused augmentations for the input imagery. A transform that mimics real-world noise would be one that simulates patchy clouds over images. This would have a similar effect to masking used in mainstream computer vision but more realistic.
Describe the solution you'd like
Clouds could be simulated by generating randomly placed patchy holes with Gaussians, then using a graduated alpha layer to simulate the Gaussian gaps in the clouds with smooth edges.
The text was updated successfully, but these errors were encountered: