Drone-PUR: From Visual Clue Discovery to Understanding and Reasoning in UAV Scenes
Paper • 2610.40801 • Published
Dataset Viewer
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code: StreamingRowsError
Exception: TypeError
Message: Couldn't cast array of type
struct<question_id: string, images: list<item: struct<path: string>>, question: string, options: struct<A: string, B: string, C: string, D: string, E: string, F: string>, correct_answer: string>
to
List({'question_id': Value('string'), 'images': List({'path': Value('string')}), 'question': Value('string'), 'correct_answer': Value('string')})
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 291, in _generate_tables
pa_table = paj.read_json(
io.BytesIO(batch), read_options=paj.ReadOptions(block_size=block_size)
)
File "pyarrow/_json.pyx", line 364, in pyarrow._json.read_json
File "pyarrow/error.pxi", line 155, in pyarrow.lib.pyarrow_internal_check_status
return check_status(status)
File "pyarrow/error.pxi", line 92, in pyarrow.lib.check_status
raise convert_status(status)
pyarrow.lib.ArrowInvalid: JSON parse error: Missing a name for object member. in row 0
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/src/services/worker/src/worker/utils.py", line 149, in get_rows_or_raise
return get_rows(
dataset=dataset,
...<4 lines>...
column_names=column_names,
)
File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
return func(*args, **kwargs)
File "/src/services/worker/src/worker/utils.py", line 129, in get_rows
rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
File "/src/services/worker/src/worker/utils.py", line 544, in safe_iter
yield from ds.decode(False) if ds.features else ds
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 3052, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2559, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2584, in _iter_arrow
for key, pa_table in self.ex_iterable._iter_arrow():
~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 558, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 440, in _iter_arrow
for key, pa_table in self.generate_tables_fn(**gen_kwags):
~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 357, in _generate_tables
yield Key(shard_idx, 0), self._cast_table(pa_table)
~~~~~~~~~~~~~~~~^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
pa_table = table_cast(pa_table, self.info.features.arrow_schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2381, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2315, in cast_table_to_schema
cast_array_to_feature(
~~~~~~~~~~~~~~~~~~~~~^
table[name] if name in table_column_names else pa.array([None] * len(table), type=schema.field(name).type),
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
feature,
^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 1861, in wrapper
return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])
~~~~^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2161, in cast_array_to_feature
raise TypeError(f"Couldn't cast array of type\n{_short_str(array.type)}\nto\n{_short_str(feature)}")
TypeError: Couldn't cast array of type
struct<question_id: string, images: list<item: struct<path: string>>, question: string, options: struct<A: string, B: string, C: string, D: string, E: string, F: string>, correct_answer: string>
to
List({'question_id': Value('string'), 'images': List({'path': Value('string')}), 'question': Value('string'), 'correct_answer': Value('string')})Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Drone-PUR Training Set
Paper | Code | Evaluation Benchmark
173,451 training samples across 11 tasks, provided in 14 task/subtask JSON files. The evaluation benchmark is released separately.
Images
This repository includes only 56,971 newly collected Dronemillion images used by the training set. GPS/location metadata has been removed; original pixels and dimensions are preserved.
The remaining 87,062 distinct source images are from the eight public datasets below and are not included. Download them through the VisDrone platform and the official project links, then place them under Drone-PUR/images/ as shown below. Keep the original filenames and directory structure within each dataset.
| Dataset | Referenced images | Official download page |
|---|---|---|
| VisDrone | 13,347 | VisDrone-Dataset |
| DroneVehicle | 27,339 | DroneVehicle |
| DroneCrowd | 2,897 | DroneCrowd |
| MDMT | 16,126 | MDMT |
| MDOT | 15,536 | MultiDrone |
| DroneRGBT | 5,395 | DroneRGBT |
| AnimalDrone | 4,547 | AnimalDrone |
| DroneBird | 1,875 | DroneBird |
Counts refer to unique referenced source images; RGB/TIR images and different views count separately. Follow the Drone-PUR annotation split when selecting samples.
Files and Annotations
Drone-PUR-Train/
annotations/ 14 full training JSON files
Drone-PUR/
images/
Dronemillion/DM_001/... Included images (DM_001 to DM_012)
VisDrone/ Download separately
VisDrone2019-DET-train/images/...
VisDrone2019-VID-train/sequences/...
... Other referenced official splits
DroneVehicle/ train/trainimg, train/trainimgr, val/...
DroneCrowd/ train_data/images, val_data/images
MDMT/ train/..., val/...
MDOT/ MDOT-2/train-2/..., MDOT-3/three_train/...
DroneRGBT/ Train/RGB, Train/Infrared, Test/...
AnimalDrone/ AnimalDrone_PartA/..., AnimalDrone_PartB/...
DroneBird/ train/images/...
metadata/ Image manifest, template settings, task counts
tools/
LICENSE
Each JSON contains dataset, split, task, subtask, total_samples and results. Samples contain ordered images, a question, and task-specific answers: options/correct_answer for multiple choice, Yes/No for CJ, and typed ground_truth for boxes or counts. Boxes use normalized [x1, y1, x2, y2] coordinates. See task counts.
MMR image order is RGB then TIR; MVR uses View 1 then View 2. MIG retains four-frame order and query_frame. MIG and MVR-VG provide clean reference images with template.box; the training code draws the reference boxes.
Use
All image paths are relative to the downloaded repository root, for example Drone-PUR/images/DroneCrowd/train_data/images/xxx.jpg. Set the training code's --data-root to that root and use metadata/train_sources_full.json as its source configuration. No path rewriting is needed with the directory layout above.
Optionally check image availability and create a separate local annotation copy with absolute paths:
python tools/remap_external_paths.py --data-root . --output ../Drone-PUR-Train-local --check-files
See the training guide. All task pools are full, including OR and OD.
License
Annotations and included Dronemillion images: CC BY-NC-SA 4.0, see LICENSE. External images remain subject to their source datasets' terms.
- Downloads last month
- 297