Forest Fire UAV Imagery Segmented into 1280x720 Tiles

UAV-captured images of forest fires, originally 3840x2160 pixels, were split into nine non-overlapping 1280x720 tiles. Fire location data was extracted from segmentation masks provided by the primary FLAME dataset using a connected components algorithm. The dataset was created by alarmod and last updated on the Hugging Face platform in February 2024.

Use Cases

• Train image segmentation models based on fire masks from the FLAME dataset.
• Develop object detection algorithms for fire points based on extracted coordinate and size data.
• Benchmark UAV-based remote sensing models for environmental monitoring.

Strengths

• Source images have a high resolution of 3840x2160 pixels.
• Fire location data is algorithmically extracted from segmentation masks.
• Tiles are non-overlapping, which may simplify certain analysis tasks.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The description references external files (FLAME dataset files 9 and 10) for full context.

Provenance

Source

Derived from the FLAME dataset (files 9 and 10).

Collection Method

UAV imagery was split into tiles; fire data was extracted from masks using a connected components algorithm.

Freshness

Last updated 2024-02-27 11:50:09; freshness should be verified.