VQA-Illusion/IllusionAnimals_train

IllusionAnimals — Training Set

Dataset summary

This repository contains the training split of IllusionAnimals, one of the four benchmarks introduced in Illusory VQA: Benchmarking and Enhancing Multimodal Models on Visual Illusions. It supports training models to identify animal categories embedded as visual illusions (pareidolia) in generated scenes and to recognize when no illusion is present.

The source-condition animal images were generated with SDXL-Lightning. These images were combined with English scene descriptions and transformed into illusory images using ControlNet. The training set is balanced across ten animal classes and an additional No illusion class.

Property	Value
Hugging Face repository	VQA-Illusion/IllusionAnimals_train
Official split	Train
Task	11-way illusion classification / visual question answering
Annotated examples	3,300
Image format	JPEG
Metadata file	df_data.csv
Paper	arXiv:2412.08169
Code	IllusoryVQA/IllusoryVQA

Repository structure

Path	Files	Description
ill_images/	3,300	Primary training images: 3,000 illusion-bearing examples and 300 No illusion examples.
raw_images/	3,000	SDXL-Lightning source-condition animal images used for the illusion-bearing rows. No raw counterpart is provided for the 300 No illusion rows.
df_data.csv	1	Canonical metadata and labels.

The value in image_name is the filename stem. For example, IllusionAnimals_1 corresponds to ill_images/IllusionAnimals_1.jpg.

Metadata schema

Column	Type	Description
image_name	string	Image identifier and filename stem.
Pprompt	string	Positive scene prompt used during generation.
Nprompt	string	Negative prompt. It is empty for No illusion rows.
illusion_strength	float or empty	Control strength; 2.5 for illusion-bearing rows and empty for No illusion rows.
label	string	Animal class name or no illusion.

Label mapping

Although df_data.csv stores animal names, the following numeric order matches the class lists used in the official experiment code.

Numeric ID	Class label	Value stored in df_data.csv
0	cat	cat
1	dog	dog
2	pigeon	pigeon
3	butterfly	butterfly
4	elephant	elephant
5	horse	horse
6	deer	deer
7	snake	snake
8	fish	fish
9	rooster	rooster
10	No illusion	no illusion

Download

pip install -U huggingface_hub pandas pillow

from huggingface_hub import snapshot_download

dataset_dir = snapshot_download(
    repo_id="VQA-Illusion/IllusionAnimals_train",
    repo_type="dataset",
)
print(dataset_dir)

Command-line alternative:

huggingface-cli download VQA-Illusion/IllusionAnimals_train \
  --repo-type dataset \
  --local-dir IllusionAnimals_train

Load and pair images with metadata

from pathlib import Path
import pandas as pd
from huggingface_hub import snapshot_download

root = Path(snapshot_download(
    repo_id="VQA-Illusion/IllusionAnimals_train",
    repo_type="dataset",
))
df = pd.read_csv(root / "df_data.csv", keep_default_na=False)

df["illusion_path"] = df["image_name"].map(
    lambda name: root / "ill_images" / (name + ".jpg")
)
df["raw_path"] = df["image_name"].map(
    lambda name: root / "raw_images" / (name + ".jpg")
)
df["raw_path"] = df["raw_path"].map(
    lambda path: path if path.exists() else None
)

label_to_id = {
    "cat": 0,
    "dog": 1,
    "pigeon": 2,
    "butterfly": 3,
    "elephant": 4,
    "horse": 5,
    "deer": 6,
    "snake": 7,
    "fish": 8,
    "rooster": 9,
    "no illusion": 10,
}
df["label_normalized"] = df["label"].str.strip().str.casefold()
df["label_id"] = df["label_normalized"].map(label_to_id)

assert df["label_id"].notna().all()
assert df["illusion_path"].map(Path.exists).all()

Intended use

Suitable uses include supervised illusion-aware animal classification, constrained-answer VQA, comparisons between source-condition and illusion images, and robustness studies for multimodal models. Expected answers are the ten animal names plus No illusion.

Dataset creation and safety

The authors generated source-condition animal images with SDXL-Lightning, generated diverse English scene prompts with several language models, and produced illusion images with ControlNet. Human reviewers validated dataset quality. The paper reports NSFW screening and the exclusion of flagged images from the public release.

Important usage notes

• Use df_data.csv as the authoritative index.
• Hugging Face may auto-detect top-level image folders as imagefolder classes. These folder labels are storage variants, not animal targets.
• Labels are stored as text, even though the experiment code uses the numeric ordering shown above.
• Only illusion-bearing rows have files in raw_images/.
• The benchmark primarily contains one large hidden category per image; consult the paper for full limitations.

License

This dataset repository declares the MIT license. Users should also review and comply with any applicable terms associated with upstream models and components.

Citation

@misc{rostamkhani2024illusoryvqa,
  title        = {Illusory VQA: Benchmarking and Enhancing Multimodal Models on Visual Illusions},
  author       = {Rostamkhani, Mohammadmostafa and Ansari, Baktash and Sabzevari, Hoorieh and Rahmani, Farzan and Eetemadi, Sauleh},
  year         = {2024},
  eprint       = {2412.08169},
  archivePrefix = {arXiv},
  primaryClass = {cs.CV},
  url          = {https://arxiv.org/abs/2412.08169}
}

Contact

Questions and reproducibility issues can be submitted through the official GitHub repository.