Hello, I have uploaded images on Roboflow and I have forgotten to annotate them.
Then, I have generated the dataset and it is now impossible for me to export it on a YOLO format.
Please, How to fix this issue?
It is possible for me to go back and generate annotations for all the images?
Once you generate a version of a dataset, you cannot change it, but what you can always do is generate a new version of the dataset, once you’ve annotated the images you forgot.
Also, it’s worth noting that images you don’t annotate do not automatically get included in generated versions. Most likely, they’ll remain as uncompleted annotation jobs in the Annotate tab:
It might be worth checking if they are still there. They might not be there if you marked them as null. In that case, you can find the images you are looking for by going into the Dataset tab and searching for images marked null:
Hope this helps
Thank you so much @leo . By following your tips, I have been able to generate a new version.
But , as you can see on this image, the download of the dataset doesn’t suggest me the YOLO format like an option. Please, what is wrong?
Glad to hear you were able to generate a new version of your dataset. Can you tell me which notebook you are using?
If you are using our YOLOv8 classification notebook, I’ve just noticed and created a pull request for a fix to update our notebook. It say to export using YOLO v5 PyTorch when it should say Folder Structure. If you are using that notebook, please select the folder option.
Hope this helps
I am using YOLOV7 training notebook.
When I select folder option, the file data.yaml is missing. So, I can not move on without this file.
Please @leo you can try to download the dataset on your side through this link:
You will see that they don’t suggest any YOLO format.
Good morning Leo !
Wishing you a happy week.
I am writing to please you to take into account my problem of exporting the dataset in YOLOV7 format . I strongly need this to make progress on my master’s thesis.
Thank you in advance.
Best regards!
Hello Leo!
I know what is the problem. Infact, I defined my project as a multi-label classification.
Please, is it possible to convert it to an object detection project?
Hi @Audrey_Souop
I was under the impression that you were trying to make a classification model, sorry about that. To change it into an object detection model, you should download the dataset, create a new object detection project and upload the contents there.
If you have any more issues or have trouble getting that done, feel free to let us know!
Hello Leo!
Thank you, I will do that.
Please, will I still need to annotate the images?
Unfortunately, yes. You will most likely have to annotate the files. I recommend training versions of your model in between your annotations so you can use Label Assist.
Good luck with your thesis!
Ok 
. The dataset contains more than 2000 images and I previously annotated them with Roboflow.
I am very grateful for your assistance Leo.
I’ve added two extra training credits so that you can easily train new models to help your labeling job go faster. Hope this helps.
Thank you!
Although I don’t know how to take advantage of that.
What you can do is when you’re done with labeling a portion of your dataset, generate a new version and train it.
Then, you can see how to use Label Assist to help you label images faster.
Thank you Leo, I have done it today.
