Synthetic CCTV Smoke & Fire Detection Image Data | Early Fire & Smoke | YOLO + COCO Annotations

Fire detection has to work in the first minutes, when the smoke is thin and the flames are small, and it must not alarm on every steam vent or floodlight. Real footage of early fires on security cameras is scarce. This dataset gives you that footage at volume. What's inside - Synthetic CCTV images of early-stage fires and smoke in outdoor yards, car parks, industrial estates, depots, loading bays and building perimeters. - Ceiling, wall and pole-mounted camera views, by day, at dusk and at night. - Small, distant and partly hidden fires and thin smoke as well as larger plumes, the way they look on real security cameras. - About a sixth of the images are look-alike negatives: steam vents, exhaust vapour, ground fog, irrigation mist, dust, floodlights, streetlights, amber beacons and low sun reflections, the glows and plumes that detectors confuse with fire and smoke. - Classes: person, smoke, fire. - Every person, smoke and fire region boxed, in YOLO and COCO format, with a train/val split, data.yaml and a dataset card. Volumes - 1,000-image pack available off the shelf (zip download). - Up to 20,000 images available on request, plus 100+ matching CCTV video clips (720p-1080p, 20 FPS, MP4). - Custom packs for your camera angles, sites and fire scenarios. Why synthetic - No real people or locations, so no GDPR or consent issues. - Fires you could never safely stage, at volume. - Every image reviewed for realism and label accuracy before release. Use cases: fire and smoke detect