The dataset viewer is not available for this subset.
Exception: SplitsNotFoundError
Message: The split names could not be parsed from the dataset config.
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.12/site-packages/datasets/inspect.py", line 289, in get_dataset_config_info
for split_generator in builder._split_generators(
^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/packaged_modules/folder_based_builder/folder_based_builder.py", line 133, in _split_generators
analyze(archives, downloaded_dirs, split_name)
File "/usr/local/lib/python3.12/site-packages/datasets/packaged_modules/folder_based_builder/folder_based_builder.py", line 108, in analyze
_, downloaded_dir_file_ext = os.path.splitext(downloaded_dir_file)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/utils/file_utils.py", line 710, in xsplitext
if is_local_path(a):
^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/utils/file_utils.py", line 72, in is_local_path
return urlparse(url_or_filename).scheme == "" or os.path.ismount(urlparse(url_or_filename).scheme + ":/")
^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/urllib/parse.py", line 395, in urlparse
splitresult = urlsplit(url, scheme, allow_fragments)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/urllib/parse.py", line 516, in urlsplit
_check_bracketed_netloc(netloc)
File "/usr/local/lib/python3.12/urllib/parse.py", line 447, in _check_bracketed_netloc
raise ValueError("Invalid IPv6 URL")
ValueError: Invalid IPv6 URL
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 65, in compute_split_names_from_streaming_response
for split in get_dataset_split_names(
^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/inspect.py", line 343, in get_dataset_split_names
info = get_dataset_config_info(
^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/inspect.py", line 294, in get_dataset_config_info
raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.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.
HVPD: Harmful Visual Plots Dataset
Overview
HVPD (Harmful Visual Plots Dataset) is a large-scale image dataset designed to support research on the automatic detection of deceptive data visualizations. The dataset focuses exclusively on visual design manipulations, cases where charts mislead through construction choices rather than through accompanying text.
HVPD is intended for:
- training and evaluating computer vision models for deceptive chart detection,
- benchmarking chart understanding and robustness,
- studying visual misinformation and deceptive design practices,
- educational and methodological research in data visualization literacy.
The dataset is designed to be scalable and extensible, enabling the addition of new deception categories in future work.
Visual Deception Categories
HVPD currently covers five common visual manipulation techniques:
Truncated Y Axis
Suppression of the zero baseline, exaggerating perceived differences and distorting effect size interpretation.Two Y Axes
Use of multiple vertical scales on a shared horizontal axis to suggest unsupported correlations.3D Pie Charts
Use of three-dimensional perspective and tilt, distorting perceived slice sizes and impairing accurate comparison.Bright Colours and No Scale
Decorative or vivid color schemes combined with missing or unclear scales, diverting attention from quantitative interpretation.Trend Hiding by Stacking
Stacking multiple series to create shifting baselines that obscure individual trends. This manipulation has not been explicitly formalized in prior datasets.
Where applicable, deceptive charts are paired with non-deceptive counterparts rendered from the same underlying data.
Dataset Composition
HVPD consists of two complementary subsets, each supporting a different stage of automated detection.
Chart Identification Subset (CIS)
The Chart Identification Subset (CIS) supports binary image classification into:
chartjust_image(non-chart images that may visually resemble charts)
Purpose
CIS enables training a front-end model that filters images before deceptive visualization analysis.
Size
- 58,501 images total
- 20,738 charts
- 37,763 non-chart images
Sources
- Public chart datasets (e.g., Kaggle)
- Official government chart repositories (USDA ERS)
- Web-scraped images retrieved using chart-like and chart-confusable query labels
All images were manually verified and labeled.
Deceptive Visualization Classification Subset (DVCS)
The Deceptive Visualization Classification Subset (DVCS) contains synthetically generated charts exhibiting controlled visual manipulations.
Size
- 24,480 chart images
- 12,050 deceptive
- 12,430 non-deceptive
Covered Classes (deceptive; non-deceptve; total)
- Truncated Y Axis (4,200; 4,200; 8,400)
- Two Y Axes (2,046; 512; 2,558)
- 3D Pie Charts (1,016; 3,072; 4,088)
- Bright Colours and No Scale (2,560; 0; 2,560)
- Trend Hiding by Stacking (2,228; 4,646; 6,874)
Non-deceptive counterparts were generated wherever feasible by correcting the manipulation while preserving the underlying data.
Data Generation Process
DVCS charts were generated in two steps:
- Numerical data was programmatically generated and stored in CSV format.
- Charts were rendered using Python visualization libraries (e.g., Matplotlib, Seaborn) and custom scripts, with randomized stylistic parameters such as colors, backgrounds, and layouts.
All generation tools were self-developed, ensuring full control over visual parameters.
Annotation and Validation
- CIS images were manually labeled.
- DVCS labels are deterministic, based on controlled chart generation.
- Dataset structure supports direct comparison between deceptive and non-deceptive encodings while controlling for data content.
Intended Use
HVPD is suitable for:
- computer vision–based chart classification,
- automated detection of deceptive visual design,
- benchmarking visual misinformation detection systems,
- educational tools for visualization literacy.
The dataset focuses exclusively on visual design flaws and does not include textual narratives.
Availability
HVPD is publicly available to support transparency, reproducibility, and further research. Researchers are encouraged to:
- compare detection models,
- extend the dataset with new deception categories,
- reuse it in educational or methodological studies.
- Downloads last month
- 15