Datasets:

Modalities:
Image
Text
Formats:
parquet
Languages:
English
Size:
< 1K
DOI:
Libraries:
Datasets
pandas
License:
leon-se's picture
Update README.md
df3ca44 verified
metadata
dataset_info:
  features:
    - name: image
      dtype: image
    - name: filename
      dtype: string
    - name: prompt
      dtype: string
    - name: response_schema
      struct:
        - name: properties
          struct:
            - name: confirm_uncontrolled_forest_fire
              struct:
                - name: enum
                  sequence: string
                - name: type
                  dtype: string
            - name: fire_hotspots
              struct:
                - name: enum
                  sequence: string
                - name: type
                  dtype: string
            - name: fire_intensity
              struct:
                - name: enum
                  sequence: string
                - name: type
                  dtype: string
            - name: fire_size
              struct:
                - name: enum
                  sequence: string
                - name: type
                  dtype: string
            - name: fire_state
              struct:
                - name: enum
                  sequence: string
                - name: type
                  dtype: string
            - name: fire_type
              struct:
                - name: enum
                  sequence: string
                - name: type
                  dtype: string
            - name: forest_fire_flames_visible
              struct:
                - name: enum
                  sequence: string
                - name: type
                  dtype: string
            - name: forest_fire_smoke_visible
              struct:
                - name: enum
                  sequence: string
                - name: type
                  dtype: string
            - name: infrastructure_nearby
              struct:
                - name: enum
                  sequence: string
                - name: type
                  dtype: string
            - name: people_nearby
              struct:
                - name: enum
                  sequence: string
                - name: type
                  dtype: string
            - name: tree_vitality
              struct:
                - name: enum
                  sequence: string
                - name: type
                  dtype: string
        - name: required
          sequence: string
        - name: type
          dtype: string
    - name: gt_answer
      dtype: string
    - name: gt_dict
      struct:
        - name: confirm_uncontrolled_forest_fire
          dtype: string
        - name: fire_hotspots
          dtype: string
        - name: fire_intensity
          dtype: string
        - name: fire_size
          dtype: string
        - name: fire_state
          dtype: string
        - name: fire_type
          dtype: string
        - name: forest_fire_flames_visible
          dtype: string
        - name: forest_fire_smoke_visible
          dtype: string
        - name: infrastructure_nearby
          dtype: string
        - name: people_nearby
          dtype: string
        - name: tree_vitality
          dtype: string
  splits:
    - name: train
      num_bytes: 48192256
      num_examples: 301
  download_size: 46089556
  dataset_size: 48192256
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
license: apache-2.0
language:
  - en
tags:
  - climate

Dataset Card: ForestFireInsights-Eval

Model Description

This dataset contains data from the Center for Wildfire Research at University of Split (Website, Paper) and original UAV images from the student team Evolonic at FAU Erlangen and Fraunhofer IISB. We thank Lara Schindhelm, Sebastian Wiederhold and Adrian Sauer for gathering images with various drones! The dataset is formatted for use with our ForestFireVLMs (7B and 3B)

Citations

This model is associated with research currently under peer review with MDPI. Please cite our paper when using this model:

[Citation will be added when the paper is published]