By using AWS re:Post, you agree to the Terms of Use
/How to load annotation from local computer in Amazon rekognition custom label service/

How to load annotation from local computer in Amazon rekognition custom label service


Hello Everyone,

I am creating object detection model using Amazon rekognition custom label service. I have already uploaded the image data in s3 and attached it to rekognition project. In Label part, I can select only 9 Images at a time After saving the annotation it takes me to 1st page from there I have to go to 30 or 40 page wherever I last saved the annotation. Doing annotation on that tools is very time consuming.

So can we use the locally annotated images in amazon rekognition service ? Is there any way to upload the json file of the annotation and we can directly train the model.

Any help will be appreciated.

Thank you Viraj Hapaliya

1 Answers

You can upload the label as a SageMaker Ground Truth format manifest file. It's a JSON lines format and you can find some example here: . You might have to transform the format if you local annotation tool generates a different format.

Once you have the manifest file you can import it using the Console or SDK, see

answered 18 days ago
  • I was able to load my local annotation into sagemaker by making changes in the json. But How do I download the manifest file of the annotation which I have done in amazon rekognition custom labels.

    I have checked my S3 bucket but I am not to find output.menifest in that.

    And One more question I have already added the sagemaker image files in amazon rekognition service. Now I want to update manifest file in rekognition service.

  • Hi Viraj, You could export the JSON Lines information of your dataset using this API: I don't see a easy way to update the manifest file in-place. But I believe you can upload a new manifest file to a S3 bucket, keeping all the images untouched, and then delete and recreate the dataset in Rekognition using the new manifest file.

You are not logged in. Log in to post an answer.

A good answer clearly answers the question and provides constructive feedback and encourages professional growth in the question asker.

Guidelines for Answering Questions