Datasets:

---

chcorbi

/

helvipad

like 9

I refreshed the Viewer and there is a new error:

PIL.UnidentifiedImageError: cannot identify image file <_io.BytesIO object at 0x7f7789051680>

it looks like some of the images can't be loaded properly

Hi @lhoestq ,

Thank you for your help. I have updated the configuration in README to only show images_bottom:

 configs:
  - config_name: default
    data_files:
      - split: train
        path: train/images_bottom/**
      - split: val
        path: val/images_bottom/**
      - split: test
        path: test/images_bottom/**
    default: true

Would i be possible to refresh the Viewer again as the error code is now back to
Error code: JobManagerCrashedError

Hi ! I refreshed it and this time it seems there is a bug with the Viewer and it doesn't show the images_bottom images as requested. It should be fixed soon, once we have updated the Viewer's backend to the latest datasets library version

Hi @lhoestq , any update on that subject?

I relaunched the Viewer and unfortunately the latest datasets version didn't help.

I've looked more closely to your dataset and it seems you are defining features in the README.md YAML header.
However the metadata that link the images together are missing, you need e.g. a metadata.csv file where you link which depth map goes with with other image, e.g.

images_top_file_name,images_bottom_file_name,depth_maps_file_name,disparity_maps_file_name,depth_maps_augmented_file_name,disparity_maps_augmented_file_name
...
...

and containing the path of each image in the repository

you can find more docs at https://huggingface.co/docs/hub/datasets-image and https://huggingface.co/docs/datasets/video_dataset