Error message in AutoML: "This dataset doesn't have annotations."

[julienpritwtpe]

Dear Community,

This is what I did:

• I created a dataset with 6 folders named AO1, AO2, AO3, AO4, AO5, and AO6.

• Each folder has between 200 and 400 one-page pdfs

• I zipped the 6 folders into one zip file

• I went to AutoML, added a "New Dataset", gave a name, selected "Multi-label Classification" (this is what I need), and pressed "Create dataset
• I then uploaded the zip file, selected a destination on Google Cloud Storage and pressed "Import"
• While it took some time, the importation was successful
• Unfortunately, when I want to "Start training", I am not able to because I have the following error message: "This dataset doesn't have annotations."

I do not see what I am doing wrong.

Any suggestion?

Thanks.

Julien

[ShrirajPatil]

The error message "This dataset doesn't have annotations" means that AutoML requires labels or tags assigned to each image in your dataset for training. In the context of multi-label classification, each image should have associated labels that specify its categories or classes.So what you can do is  Upload annotation file along with your images to AutoML. This step ensures that each image is associated with the relevant labels.

[julienpritwtpe]

Hi Shriraj,

Thanks for your feedback.

Note that our dataset is not composed of images. We have 1000+ one-page PDFs. Each PDF has text in it; the PDFs are categoried in 6 different folders. Some PDFs are present in multiple folders, hence our choice of Multi-label Classification.

You wrote: "So what you can do is  Upload annotation file along with your images to AutoML"

1. Do you mean like a CSV file in which we have the names of the pdfs and the categories they are associated to?
2. Do you have an example of such a file?

Thanks.

Julien