Dataset import error (COCO) - image not found

[leoll2]

Steps to reproduce

1. Download and import this dataset as new project: https://storage.geti.intel.com/test-data/e2e-bdd/datasets/bounding_box-datumaro-small.zip
   This is one of the sample COCO datasets used in Geti 2.x e2e tests

Expected

The project is successfully created and populated with the dataset.

Actual

The import fails with the following error:

  File "/home/leoll2/training_extensions/application/backend/app/execution/dataset_import/prepare.py", line 68, in convert_archive
    export_dataset(dataset, output_path=archive_path.parent / "dataset", as_zip=False)
    │              │                    │            └ <property object at 0x74c0fa8d5ad0>
    │              │                    └ PosixPath('data/staged_datasets/13fe3138-a4b0-44a6-92d9-5aa3b75becc8/dataset.zip')
    │              └ <datumaro.experimental.dataset.Dataset object at 0x74bf800970e0>
    └ <function export_dataset at 0x74bfc010ea20>

  File "/home/leoll2/training_extensions/application/backend/.venv/lib/python3.13/site-packages/datumaro/experimental/export_import.py", line 677, in export_dataset
    images_paths = _export_images_from_dataset(dataset, images_dir, ignore_missing_media)
                   │                           │        │           └ False
                   │                           │        └ PosixPath('data/staged_datasets/13fe3138-a4b0-44a6-92d9-5aa3b75becc8/dataset/images')
                   │                           └ <datumaro.experimental.dataset.Dataset object at 0x74bf800970e0>
                   └ <function _export_images_from_dataset at 0x74bfc010e480>
  File "/home/leoll2/training_extensions/application/backend/.venv/lib/python3.13/site-packages/datumaro/experimental/export_import.py", line 398, in _export_images_from_dataset
    image_paths[field_name] = _export_single_field(dataset, field_name, field, output_dir, ignore_missing_media)
    │           │             │                    │        │           │      │           └ False
    │           │             │                    │        │           │      └ PosixPath('data/staged_datasets/13fe3138-a4b0-44a6-92d9-5aa3b75becc8/dataset/images')
    │           │             │                    │        │           └ ImagePathField(semantic='default', format='RGB', channels_first=False, dtype=String)
    │           │             │                    │        └ 'image'
    │           │             │                    └ <datumaro.experimental.dataset.Dataset object at 0x74bf800970e0>
    │           │             └ <function _export_single_field at 0x74bfc010e3e0>
    │           └ 'image'
    └ {}
  File "/home/leoll2/training_extensions/application/backend/.venv/lib/python3.13/site-packages/datumaro/experimental/export_import.py", line 358, in _export_single_field
    raise ValueError(

ValueError: Value was set for field image, index 0, value data/staged_datasets/13fe3138-a4b0-44a6-92d9-5aa3b75becc8/dataset_import/images/Trailcam footage of a Pine Marten and wild Fallow Deer_frame_1170.jpg, but no image could be obtained (ImagePathField/MediaPathField) or no image could be generated and saved (Image/MaskCallableField).

[AlbertvanHouten]

Should be fixed by open-edge-platform/datumaro#2069

[itallix]

Should be fixed by open-edge-platform/datumaro#2069

Duplicates #5763
I've tried to reproduce it in #5769 with the latest datumaro, getting ColumnNotFoundError when filtering dataset by labels in import_dataset_as_new_project job:

File "/Users/vkarnius/Developer/training_extensions/application/backend/.venv/lib/python3.13/site-packages/datumaro/experimental/filtering/label_filter.py", line 259, in _filter_list_label
    result = df.with_columns(columns_to_update)
             │  │            └ [<Expr ['col("label").list.eval(.when(e…'] at 0x34309E050>, <Expr ['as_struct("bboxes", col("label…'] at 0x34309E150>]
             │  └ <function DataFrame.with_columns at 0x157d91e40>
             └ shape: (47, 6)
               ┌────────────┬─────────┬─────────────────────────────────┬──────────┬────────────────────┬────────────────────...
  File "/Users/vkarnius/Developer/training_extensions/application/backend/.venv/lib/python3.13/site-packages/polars/dataframe/frame.py", line 10335, in with_columns
    .collect(optimizations=QueryOptFlags._eager())
                           │             └ <staticmethod(<function QueryOptFlags._eager at 0x157cd7740>)>
                           └ <class 'polars.lazyframe.opt_flags.QueryOptFlags'>
  File "/Users/vkarnius/Developer/training_extensions/application/backend/.venv/lib/python3.13/site-packages/polars/_utils/deprecation.py", line 97, in wrapper
    return function(*args, **kwargs)
           │         │       └ {'optimizations': <polars.lazyframe.opt_flags.QueryOptFlags object at 0x343072200>}
           │         └ (<LazyFrame at 0x343072150>,)
           └ <function LazyFrame.collect at 0x157d06340>
  File "/Users/vkarnius/Developer/training_extensions/application/backend/.venv/lib/python3.13/site-packages/polars/lazyframe/opt_flags.py", line 324, in wrapper
    return function(*args, **kwargs)
           │         │       └ {'optimizations': <polars.lazyframe.opt_flags.QueryOptFlags object at 0x344427930>}
           │         └ (<LazyFrame at 0x343072150>,)
           └ <function LazyFrame.collect at 0x157d062a0>
  File "/Users/vkarnius/Developer/training_extensions/application/backend/.venv/lib/python3.13/site-packages/polars/lazyframe/frame.py", line 2429, in collect
    return wrap_df(ldf.collect(engine, callback))
           │       │   │       │       └ None
           │       │   │       └ 'auto'
           │       │   └ <method 'collect' of 'builtins.PyLazyFrame' objects>
           │       └ <builtins.PyLazyFrame object at 0x343392250>
           └ <function wrap_df at 0x15765bba0>

polars.exceptions.ColumnNotFoundError: unable to find column "label"; valid columns: ["image_info", "subset", "bboxes", "labels", "bboxes_canvas_size", "image_path"]