baizhanquan/FireDetectionDataset-catalog

FireDetectionDataset Catalog

This repository is a metadata index. It does not contain training images.

Hugging Face dataset repository IDs support only namespace/repository-name, so the logical hierarchy is encoded with the naming rule:

FireDetectionDataset-<category>-<subcategory>-<source>

Logical taxonomy

FireDetectionDataset/
|-- flame/
|   |-- indoor/
|   |   |-- candle_flame
|   |   |-- lighter_flame
|   |   |-- match_flame
|   |   |-- gas_burner
|   |-- outdoor/small_flame
|   |-- forest
|   `-- vehicle
|-- spark/
|   |-- welding
|   |-- grinding
|   `-- electrical
`-- negative/
    |-- indoor/flame_like
    |-- lighting
    |-- reflection
    |-- sunset
    |-- welding_arc
    |-- fog_smoke
    `-- exhaust

Current datasets

Taxonomy path	Managed dataset	Status
flame/forest	FireDetectionDataset-flame-forest-flameye-wildfire	Candidate accepted for flame label review; CC BY 4.0
flame/outdoor/small_flame	FireDetectionDataset-flame-outdoor-small-flame-vn	Small-object candidate; license review required
flame/indoor/lighter_flame	FireDetectionDataset-flame-indoor-lighter-flame	One boxed candidate; license review required
flame/indoor/candle_flame	FireDetectionDataset-flame-indoor-candle-flame	One boxed candidate; license review required
flame/indoor/match_flame	FireDetectionDataset-flame-indoor-match-flame	Three independent boxed candidates; license review required
flame/indoor/gas_burner	FireDetectionDataset-flame-indoor-gas-burner	Collection target; no qualified source found
negative/indoor/flame_like	FireDetectionDataset-negative-indoor-flame-like	Collection target; no accepted samples yet

| lame/indoor/{candle,lighter,match} | FireDetectionDataset-flame-indoor-characteristic-flame-barman | 8,452-image YOLO source mirrored; CC BY-NC-SA 4.0; curation pending |

Admission rule

A repository being listed here does not mean its samples are ready for training. Sources must pass license verification, label normalization, exact and perceptual deduplication, scene-group split review, and human label sampling before they enter flame_clean_v1.