Datasets:

RISys-Lab
/

SAFIRE_83K

dataset_info:
  - config_name: default
    features:
      - name: image
        dtype: image
      - name: file_name
        dtype: string
      - name: text
        dtype: string
      - name: scenario
        dtype: string
    splits:
      - name: train
        num_examples: 82842
    download_size: 2976494482
    dataset_size: 2976494482
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
task_categories:
  - image-to-text
language:
  - en
tags:
  - image-captioning
  - fire
  - smoke
  - safety
  - computer-vision
size_categories:
  - 10K<n<100K
arxiv: 2609.07823

SAFIRE 83K Captioned Images

SAFIRE 83K is an image-caption dataset covering 20 fire-smoke relevant scenarios.

Each row contains the embedded image, its English caption, the original relative file name, and a normalized scenario label.

Usage

from datasets import load_dataset

ds = load_dataset("RISys-Lab/SAFIRE_83K", split="train")
print(ds[0]["image"])
print(ds[0]["file_name"])
print(ds[0]["text"])
print(ds[0]["scenario"])

Fields

image: embedded image bytes decoded by Hugging Face Datasets.
file_name: original relative path from the ImageFolder version.
text: English image caption.
scenario: normalized scenario label.

Statistics

Images: 82,842
Scenarios: 20
Parquet shards: 6

Scenario	Images
Aerospace_Operation	3,839
Barbecue	6,187
Campfire_Bonfire	5,560
Candle	4,043
Explosion	3,924
Firework	6,105
Flare_Stack	2,298
Forest_Scene	5,151
Gas_Stove	5,015
Incense	4,970
Metal_Forging	2,963
Meteor	2,636
Open_Grassland	4,912
Residential_Fire	4,280
Sky_Lantern_Hot_Balloon	1,510
Smoking	4,000
Torch	2,922
Vehicle_Fire	1,616
Volcano	5,609
Waste_Disposal_Scene	5,302