Dataset Smoke Detection

This repository provides an annotated dataset for smoke-related sensing acquired with low-cost BME688 and BME690 gas sensors during controlled outdoor combustion of Mediterranean vegetation. The dataset was generated using the Bosch Sensortec development kit and includes acquisitions performed under 17 heater profiles, enabling heater-aware analysis of metal-oxide gas-sensor responses.
The archive contains 20 CSV files organized into five acquisition nodes, for a total of 131,370 annotated samples and approximately 51.6 hours of recordings. Each file stores chronological measurements together with environmental variables, heater-cycle descriptors, temporal references, and sample-level annotations.
The dataset was collected in a fully outdoor environment under natural ventilation and light wind conditions. Smoke was generated by controlled combustion of olive and pine trees, Mediterranean scrub shrubs, and pine needles. Sensing nodes were placed approximately 100 m from the combustion source to evaluate smoke detectability under realistic dispersion conditions.
This dataset may support reproducible studies on smoke detection, heater-profile comparison, calibration transfer, cross-node benchmarking, lightweight edge-AI pipelines, and temporal analysis of smoke-related events.