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
Currently there is no indication of which model and version of that model were used to produce autoclass results that are downloadable from the dashboard. Do we want to consider showing this information if it's available? To that end, is there a standard attribute in the autoclass h5 files to convey this information? I see model_id in autoclass results that we manage but I'm not sure if that's standard.
The text was updated successfully, but these errors were encountered:
Useful feature, will require investigation as to what kind of provenance is available from our classifier code. Note that this may require changes to, and a new release of, pyifcb, which handles class files, so I will open an issue there should that become necessary.
@ethanandrews has been building similar ability to retrieve different classifier results via HAB hub. A key for function of HAB hub is discovery/notification of new classification results so that downstream products can be precomputed for users looking at maps and class time series.
HAB hub solution includes an API to investigate what class products are available and retrieve them. Redundant paths to these types of products seems fine but better if ifcbdb had mechanism notify or easily discover newly accessed class and other products (blobs, features). If that ifcbdb capability existed, we would not be restricted to ingestion/sharing of class products stored on same hardware/network as HAB hub (as is the case currently).
Currently there is no indication of which model and version of that model were used to produce autoclass results that are downloadable from the dashboard. Do we want to consider showing this information if it's available? To that end, is there a standard attribute in the autoclass h5 files to convey this information? I see
model_id
in autoclass results that we manage but I'm not sure if that's standard.The text was updated successfully, but these errors were encountered: