Datasets:

benjaik
/

trufor-inference-outputs

Modalities:

Text

Tags:

image-forensics

image-manipulation-detection

trufor

inference

License:

Dataset card Data Studio Files Files and versions

xet

Community

Dataset Preview

Duplicate

Split (1)

train

The full dataset viewer is not available (click to read why). Only showing a preview of the rows.

The dataset generation failed because of a cast error

Error code:   DatasetGenerationCastError
Exception:    DatasetGenerationCastError
Message:      An error occurred while generating the dataset

All the data files must have the same columns, but at some point there are 1 new columns ({'class'})

This happened while the csv dataset builder was generating data using

hf://datasets/benjaik/trufor-inference-outputs/scores/IMD2020_scores.csv (at revision 408085c98f3c96c54f2d71b50e89fe2615cdc26a), ['hf://datasets/benjaik/trufor-inference-outputs@408085c98f3c96c54f2d71b50e89fe2615cdc26a/scores/CASIA_scores.csv', 'hf://datasets/benjaik/trufor-inference-outputs@408085c98f3c96c54f2d71b50e89fe2615cdc26a/scores/CocoGlide_scores.csv', 'hf://datasets/benjaik/trufor-inference-outputs@408085c98f3c96c54f2d71b50e89fe2615cdc26a/scores/IMD2020_scores.csv']

Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1848, in _prepare_split_single
                  writer.write_table(table)
                  ~~~~~~~~~~~~~~~~~~^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 765, in write_table
                  self._write_table(pa_table, writer_batch_size=writer_batch_size)
                  ~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/arrow_writer.py", line 773, in _write_table
                  pa_table = table_cast(pa_table, self._schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              path: string
              class: string
              score: double
              height: int64
              width: int64
              processed_height: int64
              processed_width: int64
              resized: bool
              -- schema metadata --
              pandas: '{"index_columns": [{"kind": "range", "name": null, "start": 0, "' + 1190
              to
              {'path': Value('string'), 'score': Value('float64'), 'height': Value('int64'), 'width': Value('int64'), 'processed_height': Value('int64'), 'processed_width': Value('int64'), 'resized': Value('bool')}
              because column names don't match
              
              During handling of the above exception, another exception occurred:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1369, in compute_config_parquet_and_info_response
                  parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
                                                                        ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      builder, max_dataset_size_bytes=max_dataset_size_bytes
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 948, in stream_convert_to_parquet
                  builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
                  ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1694, in _prepare_split
                  for job_id, done, content in self._prepare_split_single(
                                               ~~~~~~~~~~~~~~~~~~~~~~~~~~^
                      gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  ):
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1850, in _prepare_split_single
                  raise DatasetGenerationCastError.from_cast_error(
                  ...<4 lines>...
                  )
              datasets.exceptions.DatasetGenerationCastError: An error occurred while generating the dataset
              
              All the data files must have the same columns, but at some point there are 1 new columns ({'class'})
              
              This happened while the csv dataset builder was generating data using
              
              hf://datasets/benjaik/trufor-inference-outputs/scores/IMD2020_scores.csv (at revision 408085c98f3c96c54f2d71b50e89fe2615cdc26a), ['hf://datasets/benjaik/trufor-inference-outputs@408085c98f3c96c54f2d71b50e89fe2615cdc26a/scores/CASIA_scores.csv', 'hf://datasets/benjaik/trufor-inference-outputs@408085c98f3c96c54f2d71b50e89fe2615cdc26a/scores/CocoGlide_scores.csv', 'hf://datasets/benjaik/trufor-inference-outputs@408085c98f3c96c54f2d71b50e89fe2615cdc26a/scores/IMD2020_scores.csv']
              
              Please either edit the data files to have matching columns, or separate them into different configurations (see docs at https://hf.co/docs/hub/datasets-manual-configuration#multiple-configurations)

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.

path string	score float64	height int64	width int64	processed_height int64	processed_width int64	resized bool
Au/Au_ani_00001.jpg.npz	0.530551	256	384	256	384	false
Au/Au_ani_00002.jpg.npz	0.528843	256	384	256	384	false
Au/Au_ani_00003.jpg.npz	0.488104	256	384	256	384	false
Au/Au_ani_00004.jpg.npz	0.47698	256	384	256	384	false
Au/Au_ani_00005.jpg.npz	0.458881	256	384	256	384	false
Au/Au_ani_00007.jpg.npz	0.463949	256	384	256	384	false
Au/Au_ani_00008.jpg.npz	0.470008	256	384	256	384	false
Au/Au_ani_00010.jpg.npz	0.458084	256	384	256	384	false
Au/Au_ani_00011.jpg.npz	0.483732	256	384	256	384	false
Au/Au_ani_00012.jpg.npz	0.481643	256	384	256	384	false
Au/Au_ani_00013.jpg.npz	0.45944	256	384	256	384	false
Au/Au_ani_00014.jpg.npz	0.474167	256	384	256	384	false
Au/Au_ani_00015.jpg.npz	0.471534	256	384	256	384	false
Au/Au_ani_00016.jpg.npz	0.491391	256	384	256	384	false
Au/Au_ani_00017.jpg.npz	0.45235	256	384	256	384	false
Au/Au_ani_00018.jpg.npz	0.505758	256	384	256	384	false
Au/Au_ani_00019.jpg.npz	0.492414	256	384	256	384	false
Au/Au_ani_00020.jpg.npz	0.462212	256	384	256	384	false
Au/Au_ani_00021.jpg.npz	0.493478	256	384	256	384	false
Au/Au_ani_00022.jpg.npz	0.48123	256	384	256	384	false
Au/Au_ani_00023.jpg.npz	0.527805	256	384	256	384	false
Au/Au_ani_00024.jpg.npz	0.468383	256	384	256	384	false
Au/Au_ani_00026.jpg.npz	0.470992	256	384	256	384	false
Au/Au_ani_00027.jpg.npz	0.488791	256	384	256	384	false
Au/Au_ani_00028.jpg.npz	0.490293	384	256	384	256	false
Au/Au_ani_00029.jpg.npz	0.480976	256	384	256	384	false
Au/Au_ani_00030.jpg.npz	0.45987	256	384	256	384	false
Au/Au_ani_00031.jpg.npz	0.446554	256	384	256	384	false
Au/Au_ani_00032.jpg.npz	0.423125	256	384	256	384	false
Au/Au_ani_00033.jpg.npz	0.428713	256	384	256	384	false
Au/Au_ani_00034.jpg.npz	0.512075	384	256	384	256	false
Au/Au_ani_00035.jpg.npz	0.473531	256	384	256	384	false
Au/Au_ani_00036.jpg.npz	0.497012	256	384	256	384	false
Au/Au_ani_00037.jpg.npz	0.488802	256	384	256	384	false
Au/Au_ani_00038.jpg.npz	0.465623	256	384	256	384	false
Au/Au_ani_00040.jpg.npz	0.466567	256	384	256	384	false
Au/Au_ani_00042.jpg.npz	0.48319	256	384	256	384	false
Au/Au_ani_00043.jpg.npz	0.445413	256	384	256	384	false
Au/Au_ani_00044.jpg.npz	0.4736	256	384	256	384	false
Au/Au_ani_00045.jpg.npz	0.472067	384	256	384	256	false
Au/Au_ani_00046.jpg.npz	0.509868	384	256	384	256	false
Au/Au_ani_00047.jpg.npz	0.49568	256	384	256	384	false
Au/Au_ani_00048.jpg.npz	0.506141	256	384	256	384	false
Au/Au_ani_00049.jpg.npz	0.518305	256	384	256	384	false
Au/Au_ani_00050.jpg.npz	0.461791	256	384	256	384	false
Au/Au_ani_00051.jpg.npz	0.476914	256	384	256	384	false
Au/Au_ani_00052.jpg.npz	0.46404	256	384	256	384	false
Au/Au_ani_00053.jpg.npz	0.502637	256	384	256	384	false
Au/Au_ani_00054.jpg.npz	0.502827	384	256	384	256	false
Au/Au_ani_00055.jpg.npz	0.487367	256	384	256	384	false
Au/Au_ani_00056.jpg.npz	0.524252	256	384	256	384	false
Au/Au_ani_00057.jpg.npz	0.512472	256	384	256	384	false
Au/Au_ani_00058.jpg.npz	0.396126	256	384	256	384	false
Au/Au_ani_00059.jpg.npz	0.495032	256	384	256	384	false
Au/Au_ani_00060.jpg.npz	0.476294	256	384	256	384	false
Au/Au_ani_00061.jpg.npz	0.457666	256	384	256	384	false
Au/Au_ani_00062.jpg.npz	0.451536	256	384	256	384	false
Au/Au_ani_00063.jpg.npz	0.458298	256	384	256	384	false
Au/Au_ani_00064.jpg.npz	0.467644	384	256	384	256	false
Au/Au_ani_00066.jpg.npz	0.474187	256	384	256	384	false
Au/Au_ani_00067.jpg.npz	0.440221	256	384	256	384	false
Au/Au_ani_00068.jpg.npz	0.456153	256	384	256	384	false
Au/Au_ani_00069.jpg.npz	0.406886	256	384	256	384	false
Au/Au_ani_00070.jpg.npz	0.46247	256	384	256	384	false
Au/Au_ani_00071.jpg.npz	0.469112	384	256	384	256	false
Au/Au_ani_00072.jpg.npz	0.497193	384	256	384	256	false
Au/Au_ani_00073.jpg.npz	0.489889	384	256	384	256	false
Au/Au_ani_00074.jpg.npz	0.4561	384	256	384	256	false
Au/Au_ani_00075.jpg.npz	0.430131	256	384	256	384	false
Au/Au_ani_00076.jpg.npz	0.444486	256	384	256	384	false
Au/Au_ani_00077.jpg.npz	0.477155	500	334	500	334	false
Au/Au_ani_00078.jpg.npz	0.439311	256	384	256	384	false
Au/Au_ani_00079.jpg.npz	0.468942	256	384	256	384	false
Au/Au_ani_00080.jpg.npz	0.478721	256	384	256	384	false
Au/Au_ani_00081.jpg.npz	0.501184	256	384	256	384	false
Au/Au_ani_00082.jpg.npz	0.453292	256	384	256	384	false
Au/Au_ani_00083.jpg.npz	0.456652	384	256	384	256	false
Au/Au_ani_00084.jpg.npz	0.522329	256	384	256	384	false
Au/Au_ani_00085.jpg.npz	0.494116	256	384	256	384	false
Au/Au_ani_00086.jpg.npz	0.480119	256	384	256	384	false
Au/Au_ani_00087.jpg.npz	0.453871	256	384	256	384	false
Au/Au_ani_00088.jpg.npz	0.498964	384	256	384	256	false
Au/Au_ani_00089.jpg.npz	0.449705	256	384	256	384	false
Au/Au_ani_00090.jpg.npz	0.509597	384	256	384	256	false
Au/Au_ani_00091.jpg.npz	0.491489	256	384	256	384	false
Au/Au_ani_00092.jpg.npz	0.483721	384	256	384	256	false
Au/Au_ani_00093.jpg.npz	0.466882	256	384	256	384	false
Au/Au_ani_00094.jpg.npz	0.473088	256	384	256	384	false
Au/Au_ani_00095.jpg.npz	0.459941	256	384	256	384	false
Au/Au_ani_00096.jpg.npz	0.465961	256	384	256	384	false
Au/Au_ani_00097.jpg.npz	0.466884	256	384	256	384	false
Au/Au_ani_00098.jpg.npz	0.479882	256	384	256	384	false
Au/Au_ani_00099.jpg.npz	0.454138	256	384	256	384	false
Au/Au_ani_00100.jpg.npz	0.465843	256	384	256	384	false
Au/Au_ani_10001.jpg.npz	0.472239	536	800	536	800	false
Au/Au_ani_10002.jpg.npz	0.491373	536	800	536	800	false
Au/Au_ani_10003.jpg.npz	0.443374	261	400	261	400	false
Au/Au_ani_10004.jpg.npz	0.487499	480	640	480	640	false
Au/Au_ani_10005.jpg.npz	0.494066	480	640	480	640	false
Au/Au_ani_10102.jpg.npz	0.458144	180	240	180	240	false

End of preview.

TruFor Phase-3 Inference Outputs

This dataset repository contains predictions produced by the public benjaik/trufor-ph2 phase-3 best checkpoint. It contains predictions only; no source dataset images or reference masks are redistributed.

Dataset	Images	Archive	Mean detection score	Images resized for inference
IMD2020	2,423	`archives/IMD2020_output.tar`	0.538848	858
CASIA 2.0 revised	11,996	`archives/CASIA_output.tar`	0.517168	0
CocoGlide	1,024	`archives/CocoGlide_output.tar`	0.469016	0

The scores/ directory provides lightweight CSV indexes containing the relative output path, detection score, original dimensions, processed dimensions, and whether resizing was required. inference_summary.json contains aggregate statistics. SHA-256 hashes for the large archives are in SHA256SUMS.

The IMD2020 archive now includes 2,009 tampered images and 414 authentic *_orig images. Its score CSV includes a class column so the two groups can be evaluated separately. The mean detection scores are 0.545677 for the tampered group and 0.505709 for the authentic group.

Each .npz prediction contains:

map: manipulation-localization probability map;
conf: confidence map;
score: image-level manipulation score;
imgsize: original image height and width;
processed_imgsize: dimensions used by the model.

Images with a dimension above 1024 pixels were proportionally resized to fit the 11 GiB inference GPU. Their map and conf arrays were then restored to the original image dimensions with bilinear interpolation.

Download

Download only the small score tables:

hf download benjaik/trufor-inference-outputs \
  --repo-type dataset --include "scores/*" --local-dir trufor-scores

Download and extract one complete output archive:

hf download benjaik/trufor-inference-outputs \
  archives/IMD2020_output.tar --repo-type dataset --local-dir .
tar -xf archives/IMD2020_output.tar

The upstream TruFor license permits informational and nonprofit use and imposes additional restrictions. Review the licenses in the model repository and the licenses of the original datasets before use or redistribution.

Downloads last month: 67

Total file size:

30.3 GB

TruFor Phase-3 Inference Outputs

Contents

Download