There are use cases like radiology where a radiologist would like to images from axial view, sagital view, coronal view in a single labeling project.

To create a dataset with multiple folders first create an archive (zip or tar) file that preserves the folder structure including the image file.

  1. Create an archive (zip or tar) containing folders and files.
  2. Login to https://app.trainingdata.io.
  3. Under Datasets select "Add Dataset".
  4. Select "Remote hosted dataset" and upload the above mentioned archive file.
  5. Inspect the newly created dataset.

External Cloud Storage & Multiple Folders

If the dataset is hosted on an external cloud storage, multiple folders are supported by specifying the folder in the first column as shown below. Folders are "dcm" and "png".

Multiple folder in CSV file

File Formats

TrainingData.io application supports the following image file formats.

PNG, JPEG, BMP, DICOM (natively), NIFTI (natively), MOV (H.264, H.265, VP8/9 natively), MP4 (H.264, H.265, VP8/9 natively), TAR, ZIP, GZ, CSV