Because there can be multiple annotations per image (e.g. multiple faces or other objects), and not all machine annotations may have a matching human annotation, we need a matching process between the two.
This will allow a user to switch directly between human and machine annotation for a particular object/face.
Initially this could be as simple as finding the nearest centroid.
Because there can be multiple annotations per image (e.g. multiple faces or other objects), and not all machine annotations may have a matching human annotation, we need a matching process between the two.
This will allow a user to switch directly between human and machine annotation for a particular object/face.
Initially this could be as simple as finding the nearest centroid.