The pipeline completes without errors but the predicted class labels are wrong or nonsensical.
Possible causes and fix:
The label-file-path in the postprocess-config block of the vvas_xinfer
config JSON might point to a label file that does not match the model being used.
Verify that the label file is correct and relevant for the model - using a
mismatched label file produces incorrect class names even when inference is
working correctly.