no labels, images labelled even though i got " Labeled dataset created - ready for distillation."

[andysingal]

Search before asking

• I have searched the Autodistill issues and found no similar bug report.

Bug

no images, labels generated even after running the code: https://colab.research.google.com/drive/1DqvrUVmnhlhrmRC_OfWI3nQqQX1GlgkN?usp=sharing
Dataset: https://www.kaggle.com/datasets/dpamgautam/grocery-dataset

import supervision as sv

dataset = sv.DetectionDataset.from_yolo(
    images_directory_path=IMAGES_DIRECTORY_PATH,
    annotations_directory_path=ANNOTATIONS_DIRECTORY_PATH,
    data_yaml_path=DATA_YAML_PATH)

len(dataset)

Output

0

Environment

Kaggle notebook

!pip install -q \
autodistill \
autodistill-grounded-sam \
autodistill-yolov8 \
supervision==0.9.0 \
roboflow 

Minimal Reproducible Example

no images, labels generated even after running the code: https://colab.research.google.com/drive/1DqvrUVmnhlhrmRC_OfWI3nQqQX1GlgkN?usp=sharing

Additional

No response

Are you willing to submit a PR?

• Yes I'd like to help by submitting a PR!

[capjamesg]

Hello there! Can I confirm your dataset is in the YOLOv5 PyTorch TXT format?

[andysingal]

Hello there! Can I confirm your dataset is in the YOLOv5 PyTorch TXT format?

Here is the dataset : https://www.kaggle.com/datasets/dpamgautam/grocery-dataset

[capjamesg]

That is a large dataset for me to download. Can you confirm the dataset type for me so I don't have to download it?

[andysingal]

That is a large dataset for me to download. Can you confirm the dataset type for me so I don't have to download it?

it is a kaggle dataset: https://www.kaggle.com/datasets/dpamgautam/grocery-dataset

originated from : https://github.com/gulvarol/grocerydataset

<img width="315" alt="Screenshot 2023-11-17 at 10 15 09 AM"

">

Let me know if this helps.Looking forward to hearing from you

[capjamesg]

Hello there! You should pass in a folder that only contains images to the .label() function. This function does not look in sub-folders. Make sure that the extension of images is equal to the extension="" value you pass into the .label() function.

For example:

base_model.label(input_folder="images", output_folder="dataset", extension=".jpg")

will label all .jpg folders in the images folder.

[andysingal]

Hello there! You should pass in a folder that only contains images to the .label() function. This function does not look in sub-folders. Make sure that the extension of images is equal to the extension="" value you pass into the .label() function.

For example:

base_model.label(input_folder="images", output_folder="dataset", extension=".jpg")

will label all .jpg folders in the images folder.

Thanks for getting back to me. So it does not work work on multiple sub folder with images classified differently? . I already defined the ontology part before using grounding Dino to label the images.

I have shared the notebook for your reference, love to hear if there are ways to do it? . I have tried on single folder of images and it works fine but would love to try on folder with multiple folders by defining the ontology.

thanks again

[capjamesg]

So it does not work work on multiple sub folder with images classified differently?

This is correct.

I have tried on single folder of images and it works fine but would love to try on folder with multiple folders by defining the ontology.

We don't have plans to support labeling data that is in multiple folders, although we encourage you to file a PR to autodistill if this is something that you would like to see in the library!

[andysingal]

Thanks for your instant reply . Love to create a PR for it. I wrote an article to spread the word about Autodistill : https://ai.plainenglish.io/unlocking-the-power-of-autonomous-image-annotation-with-autodistill-39388e72128a?sk=71e8acd82ab342c2e2dd67acefe79f23 Thanks Again !!

…

On Tue, Nov 21, 2023 at 06:49 James ***@***.***> wrote: So it does not work work on multiple sub folder with images classified differently? This is correct. I have tried on single folder of images and it works fine but would love to try on folder with multiple folders by defining the ontology. We don't have plans to support labeling data that is in multiple folders, although we encourage you to file a PR to autodistill if this is something that you would like to see in the library! — Reply to this email directly, view it on GitHub <#90 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AE4LJNJBRQNITQ63PCPNE4DYFSIMTAVCNFSM6AAAAAA7P4U3KGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQMRQG43DQNJXGY> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[capjamesg]

No worries! Thank you for writing that article. I always love seeing people use our software to make interesting projects!