My team member deleted the labeled dataset in Azure ML Studio, is there a way to recover it with the exported COCO file?

baiyang@vivacityapp.com 5

Hi community. My team member deleted the labeled dataset in Azure ML Studio, but the good thing is we have exported the COCO file and MLTable for that dataset, may I ask is there a way to recover it?

baiyang@vivacityapp.com 5 Reputation points

2025-03-05T09:45:14.6933333+00:00

I have tried with the import function, it failed without the specific reason.

I used the relative path in 'file_name' and also the absolute path in 'absolute_url' field. Both result in the same error
baiyang@vivacityapp.com 5 Reputation points

2025-03-05T09:45:59.55+00:00

any suggestions and potential solution will be appreciated, thanks.
baiyang@vivacityapp.com 5 Reputation points

2025-03-05T10:15:42.6033333+00:00

Update!

After I create a new dataset and import by the relative path, 502 error happens instead of 404
Vikram Singh 2,240 Reputation points Microsoft Employee

2025-03-06T14:03:05.6033333+00:00
Hi @******@vivacityapp.com

Welcome to the Microsoft Q&A and thank you for posting your questions here.

I understand the frustration of accidentally deleting a resource in Azure Machine Learning Studio. Fortunately, there are steps you can take to recover your data labels from the exported COCO file and re-import them into a new data labeling project.

Verify the COCO File: Ensure that the COCO file you exported contains the necessary annotations. The COCO format typically includes information about images, annotations, categories, and other metadata.

Create a New Data Labeling Project: In Azure Machine Learning Studio, create a new data labeling project. Make sure to select the appropriate project type (e.g., image classification, object detection).

Prepare the COCO File for Import: The COCO file needs to be in a format that Azure Machine Learning can recognize. If there are any discrepancies or formatting issues, you might need to adjust the file. Ensure that the bounding box coordinates and other annotations are correctly specified.

Upload the COCO File: Upload the COCO file to the new data labeling project. You can do this by navigating to the "Data" section of your project and selecting the option to import data. Choose the COCO file and follow the prompts to complete the import process.

Verify the Imported Data: After importing the COCO file, verify that the annotations have been correctly imported into the new data labeling project. You can do this by reviewing a few sample images and their corresponding annotations.

For more info, please refer to:

Manage labeling projects - Azure Machine Learning | Microsoft Learn

Create and explore Azure Machine Learning dataset with labels

Let me know if you tried the above steps and still facing the issues, will be happy to assist to you further.

Thanks
baiyang@vivacityapp.com 5 Reputation points

2025-03-06T15:40:16.72+00:00

Hi, thank you for the response. I followed your instructions and still failed to import the COCO file. Can you please check mt COCO file format? Here is the details:

Share via

My team member deleted the labeled dataset in Azure ML Studio, is there a way to recover it with the exported COCO file?

Your answer