baizhanquan/FireDetectionDataset-flame-indoor-lighter-flame

Indoor Lighter Flame Collection Target

Taxonomy path: FireDetectionDataset/flame/indoor/lighter_flame.

Status

Collection target only. No accepted training samples are present yet.

One bounding-box candidate was verified in betasecond/jimei-fire-smoke-yolo-dataset:

• Image: valid/images/lighter-6526078__340_jpg.rf.991ea102337a611306e430a975d1170b.jpg
• Label: valid/labels/lighter-6526078__340_jpg.rf.991ea102337a611306e430a975d1170b.txt
• Source class: 0=Fire
• Content review: the box covers the visible lighter flame, not the lighter body.
• Scale: normalized box width 0.171875, height 0.609375.

The upstream repository has no dataset card or declared license. The sample is therefore recorded for provenance review only and has not been copied into this repository or accepted for training.

A filename-level scan of 26 Hugging Face fire/flame detection repositories found no other repository with paired lighter image and YOLO label files. Datasets about lighter-object classification, market data, or robot grasping are intentionally excluded.

Acceptance requirements

• The lighter must be ignited and the visible flame must be identifiable.
• Required task: object detection or an annotation format convertible to bounding boxes.
• Target class: flame (class_id=0 in flame_spark_v1).
• Include indoor variations: bright/dark rooms, different backgrounds, flame sizes, occlusion, camera distance, and hand-held use.
• Keep unlit lighters as hard-negative candidates with empty labels.
• Record source, license, revision, scene group, and duplicate hash before training acceptance.

Search keywords

lighter flame detection, ignited lighter flame, cigarette lighter fire, small flame object detection, lighter flame YOLO, handheld flame detection.