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
We always normalize the data to be between 0 and 1. However, normalizing channels in multimodal data can be somewhat complicated.
For some of our examples, we have several channels of one modality, and another channel of a different modality. For example, for one dataset, we have three XRF channels (Au, Fe, and Pt) and one ptychography channel.
What we should do is normalize the three XRF channels together, and normalize the ptychography channel separately. Assuming the first three channels are XRF, that might look something like this in the code:
data[:, :, :, :3] =_normalize_data(data[:, :, :, :3]) # Normalize the three XRF channels togetherdata[:, :, :, 3] =_normalize_data(data[:, :, :, 3]) # Normalize the ptychography channel by itself
So, we need a way to group together channels by modality so we can normalize them together.
It might actually be easiest to do this from the command line, rather than adding a UI option, because it would make sense to set this initially and not change it.
But I'm not sure what this would look like. Maybe adding a command line flag where you identify channels as being part of the same modality, like `--modalities "[Au, Fe, Pt], [Recon]"
Then we'll match these to labels in the code to translate them into indices. I can provide an example.
The text was updated successfully, but these errors were encountered:
We always normalize the data to be between 0 and 1. However, normalizing channels in multimodal data can be somewhat complicated.
For some of our examples, we have several channels of one modality, and another channel of a different modality. For example, for one dataset, we have three XRF channels (Au, Fe, and Pt) and one ptychography channel.
What we should do is normalize the three XRF channels together, and normalize the ptychography channel separately. Assuming the first three channels are XRF, that might look something like this in the code:
So, we need a way to group together channels by modality so we can normalize them together.
It might actually be easiest to do this from the command line, rather than adding a UI option, because it would make sense to set this initially and not change it.
But I'm not sure what this would look like. Maybe adding a command line flag where you identify channels as being part of the same modality, like `--modalities "[Au, Fe, Pt], [Recon]"
Then we'll match these to labels in the code to translate them into indices. I can provide an example.
The text was updated successfully, but these errors were encountered: