-
Notifications
You must be signed in to change notification settings - Fork 0
Sagemaker Labelling Jobs
Ayush Agarwal edited this page Feb 13, 2023
·
1 revision
- Create a private team under AmazonSagemaker/Labellingworkforces from Sagemaker UI.
- Go to Workers Section and send invites to new workers.
- Select the private team which you created in step 1, click on workers and add workers to the team.
- Once done, follow the steps below to execute a labelling job script
Fill the config with appropriate params value Sample Config Link
- iam_role : Sagemaker Execution Role ARN
- region_name : Region in which resources needs to be created. For eg. ap-southeast-1, ap-northeast-1 etc
- private_work_team_arn : Arn of the private work team created in Step 1 above.
- pre_human_arn : Human Labeling ARN for a particular region and particular task. We have to use Object Detection ones. Link. For eg : For ap-northeast-1 and object detection task, pre_human_arn = arn:aws:lambda:ap-northeast-1:477331159723:function:PRE-BoundingBox
- acs_arn : For ACS ARN for Bounding Box Task, follow Link
- bucket_name : name of the s3 bucket where the datasets are stored
- dataset_path : relative path of the dataset to be labelled wrt the bucket
- manifest_upload_dir : s3 directory to upload the manifest file of the input dataset
- manifest_file_name : name of the input dataset manifest file
- label_list : list of labels which are to be annotated
- label_file_name : name of the json file containing label info
- label_file_upload_dir : s3 directory to upload label file