Wildfire Detection Image Dataset from US Web Searches

OpenFire is an image classification dataset for wildfire detection, collected from web searches. The dataset is categorized as containing between 1,000 and 10,000 samples and was last updated in December 2022.

Use Cases

• Train an image classification model for wildfire detection using the dataset's image modality and classification task labels.
• Analyze the characteristics of crowdsourced annotations for wildfire imagery to assess label consistency.
• Fine-tune a vision model on images sourced from US web searches to detect regional wildfire patterns.

Strengths

• Dataset size is categorized between 1,000 and 10,000 samples, providing a defined scale for model training.
• Includes both image and text modalities, allowing for multi-modal analysis.
• Annotations are crowdsourced, potentially offering diverse labeling perspectives.

Limitations

• Specific row count, column details, and image resolution are unknown, limiting precise technical assessment.
• Data is sourced from web searches, which may introduce label noise or inconsistencies typical of crowdsourced annotations.
• Geographic coverage is indicated as US, which may limit model generalizability to other regions.

Provenance

Source

pyronear

Collection Method

Collected from web searches.

Freshness

Last updated on 2022-12-11.

Geography

US