Hi Everyone,
Is there any method to create a YOLO11 compatible dataset via code or utility like Roboflow for DMD dataset?. Roboflow is not allowing to generate more than 512 images for free users. I need to combine 2 or 3 Driver Monitoring System datasets for my activity. Please share if you have any code snippet in python for the same.
Best Regards,
Venkata Rao
Maybe try vcd as it was specifically called out on the dataset repository? If that’s not what you’re looking for, maybe try searching GitHub for anything related. Entirely possible you’ll have to write something yourself.
Okay, thank you.
Hi Venkata,
Certainly! To create a YOLO11 compatible dataset from sources like the Vicomtech DMD dataset, you’ll need to convert the existing annotations into the standard YOLO .txt format. This usually involves writing a custom script (e.g., in Python) to parse the original annotation files (which might be XML, JSON, etc.), extract the bounding box coordinates and class labels for each image, normalize the coordinates, and save them into corresponding .txt files.
Once your datasets are in the correct YOLO format, combining them typically involves merging the image and label directories and carefully updating the dataset’s data.yaml file to reflect the combined paths and class names, ensuring class indices are consistent across all source datasets.
You can find detailed information on the required structure and format for training datasets in the Ultralytics documentation on datasets. While we don’t provide specific conversion scripts for every public dataset, understanding the target format described there is key. Good luck with preparing your Driver Monitoring System dataset!
Thank you for your help and I understand the procedure.
You’re welcome, Venkat! Glad to hear the procedure is clear. Feel free to reach out if you encounter further questions as you work on creating your dataset.