Export for Vision Transformer

To whom it may concern,

I am emailing to inquire about an issue I am having exporting my dataset in Roboflow. I am trying to export my labeled data from Roboflow to finetune a Vision Transformer, but I am confused on the most appropriate format for doing so. Both the Roboflow blog post and Colab notebook on fine-tuning a Vision Transformer direct the user to resize their images to 224x224 and export data as a “Folder Structure” format. Unfortunately, I cannot discern where to do either of these. Indeed, when I attempt to export my labeled dataset, I do not have the option to export to a “Folder Structure” format (see attached screenshot for current list of available options).

Might someone on staff be able to assist me in this matter? If the Folder Structure format is not available, do you have another recommended format for training with the Hugging Face Vision Transformer as described in your tutorial?

I have granted support access to my workspace if needed to help debug this issue.

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.