Aerial Lidar and Imagery for Tree Species Classification in French Forests

135,569 aerial patches of monospecific forests across 449 locations in France provide data for tree species classification. Each 50 m x 50 m patch is annotated with a single tree species label, covering a total area of 339 km². The dataset was created by IGNF and includes aerial lidar and imagery.

Use Cases

• Train a vision model on aerial imagery patches to classify monospecific tree species labels.
• Analyze spatial distribution of tree species across 40 French departments using georeferenced patch data.
• Fuse aerial lidar point cloud data with imagery to improve species classification accuracy in 50 m x 50 m forest segments.

Strengths

• Large scale with 135,569 annotated patches covering 339 km².
• Geographically diverse, sourced from 449 forests across 40 French departments.
• Focused on monospecific forests, providing clear single-species labels for classification tasks.

Limitations

• Dataset is limited to monospecific forests, not representing mixed-species or complex forest ecosystems.
• Geographic coverage is biased towards southern regions of France, limiting generalizability.
• Specific details on data balance across species labels or sensor specifications are not provided in the input.

Provenance

Source

IGNF

Collection Method

Derived from aerial lidar and aerial imagery.

Freshness

Last updated in November 2024.

Geography

449 forests located in 40 French departments, mainly in southern regions.