aug_channel_dropout — Independent spectral-channel dropout¶
Group: Augmentation · C ABI: n4m_augmentation_channel_dropout_*
Description¶
Randomly drop individual wavelength channels.
Parameters¶
Name |
Type |
Default |
|---|---|---|
|
|
|
|
|
|
|
|
|
|
|
|
API and bindings¶
C ABI (ABI 2): n4m_augmentation_channel_dropout_apply · n4m_augmentation_channel_dropout_create · n4m_augmentation_channel_dropout_destroy. Use the linked public header for the exact signature, configuration, and result handles.
Python (verified public re-export):
from n4m.augmentation.spectral import ChannelDropout
Source signature: ChannelDropout(dropout_prob: float = 0.05, mode: str | int = 'zero', rng: Optional[PCG64] = None, seed: int = 0).
R: no current source-verified entry point was found for this catalog method.
MATLAB / Octave: no current source-verified entry point was found for this catalog method.
Explanations¶
Bibliographic source¶
No canonical NIR publication defines this cellwise dropout. The exact mask and replacement rules are implementation-defined.
Mathematical principle¶
Each cell is marked independently with probability dropout_prob. Marked values are either set to zero or linearly interpolated from surviving channel indices in that row. The latter uses endpoint values beyond the first or last survivor.
Appropriate uses¶
Simulating sporadic dead pixels/channels and discouraging reliance on single wavelengths.
Limits and validation¶
Real sensor failures are often persistent or contiguous rather than IID. If too few channels survive, interpolation becomes poorly informative; zero mode is representation-dependent.
Implementation¶
Python role API n4m.augmentation.spectral.ChannelDropout; ABI 2 family n4m_augmentation_channel_dropout_{create,apply,destroy}.
The ABI-2 implementation is the n4m_augmentation_channel_dropout_* lifecycle in libn4m.
Sources and provenance¶
See also: methods index · interactive dashboard