Satellite Smoke Scene Detection Imagery

USTC_SmokeRS dataset contains satellite imagery for smoke scene detection. It was created by researchers including Rui Ba and Chen Chen for the SmokeNet project, with the dataset card last updated on HuggingFace in March 2023. The dataset is intended for research and education purposes.

Use Cases

• Train a convolutional neural network with spatial attention to classify satellite images as containing smoke or not.
• Develop and benchmark models for early wildfire detection using satellite-based smoke scene labels.
• Apply channel-wise attention mechanisms to improve feature extraction from multi-spectral satellite imagery for smoke detection.

Strengths

• Dataset is associated with a peer-reviewed publication (SmokeNet, 2019), providing methodological context.
• Includes implementations for libraries like Polars, Dask, and ML Croissant, suggesting structured accessibility.

Limitations

• Critical metadata like row count, image resolution, and sample size are unknown, hindering reproducibility.
• The license is unspecified beyond 'research/education purposes', creating uncertainty for commercial use.
• Geographic coverage and temporal range of the satellite imagery are not documented, limiting assessment of bias.

Provenance

Source

Associated with the SmokeNet research article by authors from USTC (University of Science and Technology of China).

Collection Method

Satellite imagery collection and annotation for smoke scene detection, as described in the cited paper.

Freshness

Last updated on the platform in March 2023.