Just a quick update to make mistakenness functional for fiftyone==0.7.3.
Note: I'm not sure what happened, but Detection objects don't have logits, so the simple code below was apparently broken in production because it tries to use logits. I fixed this on this branch so that the simple example below runs:
import fiftyone as fo
import fiftyone.zoo as foz
import fiftyone.brain as fob
dataset = foz.load_zoo_dataset("quickstart")
fob.compute_mistakenness(dataset, "predictions")
session = fo.launch_app(dataset)
session.view = dataset.sort_by("possible_spurious", reverse=True)
Just a quick update to make mistakenness functional for fiftyone==0.7.3.
Note: I'm not sure what happened, but
Detection
objects don't havelogits
, so the simple code below was apparently broken in production because it tries to use logits. I fixed this on this branch so that the simple example below runs: