Problems with Image Label Adjustment Job in Amazon Sagemaker Ground Truth - amazon-web-services

Question

I'm trying to create a Image Label Adjustment Job in Ground Truth and I'm having some trouble. The thing is that I have a dataset of images, in which there are pre-made bounding boxes. I have an external python script that creates the "dataset.manifest" file with the json's of each image. Here are the first four lines of that manifest file:
{"source-ref": "s3://automatic-defect-detection/LM-WNB1-M-0000126254-camera_2_0022.jpg", "bounding-box": {"image_size": [{"width": 2048, "height": 1536, "depth": 3}], "annotations": [{"class_id": 0, "width": 80, "height": 80, "top": 747, "left": 840}]}, "bounding-box-metadata": {"class-map": {"0": "KK"}, "type": "groundtruth/object-detection", "human-annotated": "yes"}}
{"source-ref": "s3://automatic-defect-detection/LM-WNB1-M-0000126259-camera_2_0028.jpg", "bounding-box": {"image_size": [{"width": 2048, "height": 1536, "depth": 3}], "annotations": [{"class_id": 0, "width": 80, "height": 80, "top": 1359, "left": 527}]}, "bounding-box-metadata": {"class-map": {"0": "KK"}, "type": "groundtruth/object-detection", "human-annotated": "yes"}}
{"source-ref": "s3://automatic-defect-detection/LM-WNB1-M-0000126256-camera_3_0006.jpg", "bounding-box": {"image_size": [{"width": 2048, "height": 1536, "depth": 3}], "annotations": [{"class_id": 3, "width": 80, "height": 80, "top": 322, "left": 1154}, {"class_id": 3, "width": 80, "height": 80, "top": 633, "left": 968}]}, "bounding-box-metadata": {"class-map": {"3": "FF"}, "type": "groundtruth/object-detection", "human-annotated": "yes"}}
{"source-ref": "s3://automatic-defect-detection/LM-WNB1-M-0000126253-camera_2_0019.jpg", "bounding-box": {"image_size": [{"width": 2048, "height": 1536, "depth": 3}], "annotations": [{"class_id": 2, "width": 80, "height": 80, "top": 428, "left": 1058}]}, "bounding-box-metadata": {"class-map": {"2": "DD"}, "type": "groundtruth/object-detection", "human-annotated": "yes"}}
Now the problem is that I'm creating private jobs in Amazon Sagemaker to try it out. I have the manifest file and the images in a S3 bucket, and it actually kinda works. So I select the input manifest, activate the "Existing-labels display options". The existing labels for the bounding boxes do not appear automatically, so I have to enter them manually (don't know why), but if I do that and try the preview before creating the adjustment job, the bounding boxes appear perfectly and I can adjust them. The thing is that, me being the only worker invited for the job, the job never apears to start working on it, and it just auto-completes. I can see later that the images are there with my pre-made bounding boxes, but the job never appears to adjust those boxes. I don't have the "Automated data labeling" option activated. Is there something missing in my manifest file?

Answer 1

There can be multiple reasons for this. First of all, the automated labeling option is not support for label adjustment and verification tasks. so thats ruled out.
It looks like you have not setup the adjustment job properly. Some things to check for:
Have you specified the Task timeout and Task expiration time? If these values are practically low, then the tasks would be expired even before somebody can pick them.
Have you checked the "I want to display existing labels from the dataset for this job." box? It should be checked for your case.
Does your existing label are fetched properly? If this is not fetched correctly, either you need to review your manifest file or you need to manually provide the label values(which i guess you are doing)
Since you are the only worker in the workforce. Do you have correct permissions to access the labeling task?
How many images you have? Have you set any minimum batch size while setting the label adjustment job?