I tried to export annotations using the python SDK but for some data row (10% of them) the “labels” list is empty but the annotation exists on Labelbox. Even using the UI export the label is empty. This is strange since it works well for most of the data rows.
export_params= {
"attachments": False,
"metadata_fields": False,
"data_row_details": True,
"project_details": False,
"label_details": True,
"performance_details": False,
"interpolated_frames": False,
}
filters= {
"global_keys": ["clip_00000583_face_00000001_start_00006469_end_00006677"]
}
export_task = project.export(params=export_params, filters=filters)
export_task.wait_till_done()
# Stream results and errors
if export_task.has_errors():
export_task.get_buffered_stream(stream_type=labelbox.StreamType.ERRORS).start(
stream_handler=lambda error: print(error))
if export_task.has_result():
export_json = [data_row.json for data_row in export_task.get_buffered_stream()]
Can you share the project ID and your version of the Labelbox SDK, please?
Have you always seen this behaviour in this project, or did it start recently?
I am using Labelbox 6.5.0, and the project ID of the example is cm04ybftt01qw070yfi0o9c2d. I haven’t experienced such a problem in the past. To give you more context, the annotator had to review the pre-annotation on videos. There is both frame classification and one global classification. Out of 199 videos, around 20 don’t contain any annotation at export time while having annotation on the Labelbox web app ( at least the global classification ).
Thanks