datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
Safe_and_Unsafe_Behaviours
Dataset Card for safe_unsafe_behaviours
This is a FiftyOne dataset with 691 samples.
Installation
If you haven't already, install FiftyOne:
pip install -U fiftyone
Usage
import fiftyone as fo
from fiftyone.utils.huggingface import load_from_hub
# Load the dataset
# Note: other available arguments include 'max_samples', etc
dataset = load_from_hub("Voxel51/Safe_and_Unsafe_Behaviours")
# Launch the App
session = fo.launch_app(dataset)
Dataset… See the full description on the dataset page: https://huggingface.co/datasets/Voxel51/Safe_and_Unsafe_Behaviours.safebooru-webp-4Mpixel
Safebooru 4M Re-encoded Dataset
This is the re-encoded dataset of deepghs/safebooru_full. And all the resized images are maintained here.
There are 5756655 images in total. The maximum ID of these images is 5974383. Last updated at 2025-08-06 08:31:53 JST.
How to Painlessly Use This
Use cheesechaser to quickly get images from this repository.
Before using this code, you have to grant the access from this gated repository. And then set your personal HuggingFace token into… See the full description on the dataset page: https://huggingface.co/datasets/deepghs/safebooru-webp-4Mpixel.t2i_safety_dataset
T2ISafety: Benchmark for Assessing Fairness, Toxicity, and Privacy in Image Generation
This dataset, T2ISafety, is a comprehensive safety benchmark designed to evaluate Text-to-Image (T2I) models across three key domains: toxicity, fairness, and bias. It provides a detailed hierarchy of 12 tasks and 44 categories, built from meticulously collected 70K prompts. Based on this taxonomy and prompt set, T2ISafety includes 68K manually annotated images, serving as a robust resource for… See the full description on the dataset page: https://huggingface.co/datasets/OpenSafetyLab/t2i_safety_dataset.driver-safety-dataset
Driver Safety Dataset
This dataset contains images and annotations for driver drowsiness detection and safety monitoring.
Dataset Details
Creation Date: 2025-05-28
Task: Computer Vision - Driver Safety
Categories: Drowsy and Normal driving conditions
Dataset Structure
Raw Data:
drowsy/: Images of drowsy driving conditions
normal/: Images of normal driving conditions
Processed Data:
dataset.csv: Processed dataset with annotations
dataset_summary.json:… See the full description on the dataset page: https://huggingface.co/datasets/ckcl/driver-safety-dataset.safemaize-v2
SafeMaize v2
We use this preliminary public-source dataset for maize screening experiments.
The export contains 46,143 distinct images, including
40,385 core classification images. Files retain their original
bytes and recorded frame-selection rules.
Core class
Images
nlb_tlb_like
20,661
healthy
14,108
faw_feeding_injury
5,616
Core screening task split
Images
train
28,269
val
4,039
calibration
4,038
test
4,039
Core primary source… See the full description on the dataset page: https://huggingface.co/datasets/sathiiii/safemaize-v2.warehouse-safety-hazard-dataset
Warehouse Safety Hazard Detection Dataset
A synthetic dataset of 913 photorealistic warehouse images captured from a drone/overhead perspective, labeled across 5 safety hazard categories. Designed for training and benchmarking vision-language models (VLMs) on industrial safety inspection tasks.
Categories
Category
Train
Test
Total
Description
spill
170
43
213
Liquid spills on warehouse floors (oil, water, chemicals)
forklift_violation
108
30
138
Unsafe… See the full description on the dataset page: https://huggingface.co/datasets/Jsohal174/warehouse-safety-hazard-dataset.michiakari-safe-route-dataset
michiakari-safe-route-dataset
夜道の安心ルート提案のために作成した、自作の比較学習用データセットです。
内容
pairwise_labels.csv
pairwise_dataset.csv
segment_dataset.csv
image_index_main.csv
sample_outputs/*.json
sample_outputs/segment_scores_tie_pairplus2.csv
規模
実画像: 63枚
区間: 15
作成方法
実地で夜道画像を収集
区間単位で整理
明るさ、見通し、人の気配、開放感、歩きやすさを参考に比較ラベルを作成
制約
地域特化
小規模
画像原本は軽量公開のため一部省略
ライセンス
MIT
warehouse-safety-hazard-dataset
Warehouse Safety Hazard Detection Dataset
A synthetic dataset of 913 photorealistic warehouse images captured from a drone/overhead perspective, labeled across 5 safety hazard categories. Designed for training and benchmarking vision-language models (VLMs) on industrial safety inspection tasks.
Categories
Category
Train
Test
Total
Description
spill
170
43
213
Liquid spills on warehouse floors (oil, water, chemicals)
forklift_violation
108
30
138… See the full description on the dataset page: https://huggingface.co/datasets/sakthizz/warehouse-safety-hazard-dataset.tool-safety-dataset
Tool Safety Dataset
Dataset Description
The Tool Safety Dataset is a specialized collection of tool images with detailed safety and usage information. It combines visual data with comprehensive metadata about various hand tools, making it valuable for both computer vision tasks and safety training applications.
Dataset Summary
Type: Image dataset with bounding boxes and detailed tool information
Size: Multiple splits (train/test/validation)
Format: Images with… See the full description on the dataset page: https://huggingface.co/datasets/akameswa/tool-safety-dataset.libero_safety_v1
LIBERO Safety
This repository contains 50-scene synthetic LIBERO safety validation sets.
v1: original LIBERO safety image set.
v2: regenerated set using the latest scene configuration and saved manual positions.
v3: same scenes and labels as v2, with a static non-colliding MuJoCo white cutting board fixture added on the table; original V2 object and fixture states are unchanged.
v4: corrected zoomed non-reference set with pickable target objects for video rollouts.
v5: corrected… See the full description on the dataset page: https://huggingface.co/datasets/saaduddinM/libero_safety_v1.Safe_and_Unsafe_Behaviours
Dataset Card for safe_unsafe_behaviours
This is a FiftyOne dataset with 691 samples.
Installation
If you haven't already, install FiftyOne:
pip install -U fiftyone
Usage
import fiftyone as fo
from fiftyone.utils.huggingface import load_from_hub
# Load the dataset
# Note: other available arguments include 'max_samples', etc
dataset = load_from_hub("Voxel51/Safe_and_Unsafe_Behaviours")
# Launch the App
session = fo.launch_app(dataset)
Dataset… See the full description on the dataset page: https://huggingface.co/datasets/yangyayaa/Safe_and_Unsafe_Behaviours.road-safety-vulnerable-road-users
Road Safety & Vulnerable Road Users Visual Dataset
Rows: 484
Build with Outerview
Geographic Feature List — browse geographic features available through Outerview
Outerview — programmable infrastructure for Earth
Developer Documentation — APIs, tools, guides, and examples
SDK — build geographic capabilities directly into applications
CLI — work with geographic search and spatial data from the terminal
MCP — connect geographic search and spatial tools to AI… See the full description on the dataset page: https://huggingface.co/datasets/Outerview/road-safety-vulnerable-road-users.kamari-safe-open-v0
Kámárí-Safe Open v0 (benchmark)
A frozen, leakage-free benchmark for African-tailored age verification. It holds manifests and
split tables, not raw images (paths, hashes, labels, skin band, quality). Use it to measure age
accuracy and, more importantly, child-safety.
Headline metric
Minor-Pass-Through Rate (MPTR) is the headline: the fraction of true minors a model passes as
adults, reported overall, at 21, and for dark + brown skin. Report MPTR alongside MAE; a… See the full description on the dataset page: https://huggingface.co/datasets/Shinzmann/kamari-safe-open-v0.safebooru_full
Safebooru Full Dataset
This is the full dataset of safebooru.org. And all the original images are maintained here.
How to Painlessly Use This
Use cheesechaser to quickly get images from this repository.
Before using this code, you have to grant the access from this gated repository. And then set your personal HuggingFace token into HF_TOKEN environment variable to give the code authorization for this repository.
from cheesechaser.datapool import SafebooruDataPool
pool =… See the full description on the dataset page: https://huggingface.co/datasets/deepghs/safebooru_full.Safe_Unsafe_Test-Understanding-output-labels-qwen
Dataset Card for Safe-Unsafe-Video_Understanding
This is a FiftyOne dataset with 40 samples.
Installation
If you haven't already, install FiftyOne:
pip install -U fiftyone
Usage
import fiftyone as fo
from fiftyone.utils.huggingface import load_from_hub
# Load the dataset
# Note: other available arguments include 'max_samples', etc
dataset = load_from_hub("pjramg/Safe_Unsafe_Test-Understanding-output-labels-qwen")
# Launch the App
session =… See the full description on the dataset page: https://huggingface.co/datasets/pjramg/Safe_Unsafe_Test-Understanding-output-labels-qwen.SafeEditBench
SafeEditBench
WARNING: This repository contains content that might be disturbing!
A benchmark for evaluating content safety detection in image editing. SafeEditBench tests whether vision-language models can correctly identify policy violations in edited images across diverse content safety policies.
Dataset
SafeEditBench contains image pairs (original + edited) annotated with safety labels and policy violation types.
Split
Images
Directory
train
901… See the full description on the dataset page: https://huggingface.co/datasets/tyodd/SafeEditBench.safegrad
SafeGrad
SafeGrad is a benchmark dataset for evaluating safety properties of text-to-image (T2I) generative models. It contains 1,083 escalation ladders across 11 safety categories, each consisting of four (prompt, image) pairs ordered by monotonically increasing severity: safe → low_risk → mid_risk → high_risk.
📄 Paper / Code: https://anonymous.4open.science/r/safegrad_dataset-DC79
⚠️ Content Warning: This dataset intentionally contains unsafe prompts and AI-generated images at… See the full description on the dataset page: https://huggingface.co/datasets/safegrad/safegrad.hse-safety-inspection-data
HSE Inspection Reports Dataset
Labeled industrial safety inspection records for training HSE compliance models.
Overview
This dataset contains structured inspection reports from industrial facility safety walkthroughs. Each record documents observed conditions, violation categories, severity levels, and recommended actions — designed for training and evaluating vision-language models on HSE compliance tasks.
Structure
File
Description
Format… See the full description on the dataset page: https://huggingface.co/datasets/davidfertube/hse-safety-inspection-data.