LFD Dataset ( fire dataset)

The LFD Dataset is a specialized collection for real-time fire detection in complex scenarios, designed to enhance model robustness for embedded systems. It comprises: Public DFS Dataset: General fire/smoke scenes with diverse environmental conditions. Novel Real-World Scenes (6,000 images): Curated to address critical challenges: Blurred fire boundaries (e.g., strong lighting interference). Smoke occlusion (dense smoke obscuring flames). Small-scale fire targets (early-stage/distant fires). Cluttered backgrounds (complex scenes with flame-like distractors). Purpose Train/evaluate fire detection models (e.g., YOLO-LF) under realistic, high-interference conditions. Improve generalization for embedded firefighting devices (drones, surveillance systems). Applications Fire prevention systems Real-time emergency response Lightweight CNN development Key Features Size: 6000+ enriched images. Diversity: Covers edge cases absent in existing datasets. Benchmark Suitability: Used to validate state-of-the-art models .