NeonTreeEvaluation: Training Data for Individual Tree Detection from Airborne Imagery

Training Data for the NeonTreeEvaluation Benchmark provides large training files for detecting individual trees from airborne imagery. The dataset includes RGB images, LiDAR tiles, 426-band hyperspectral files, and 1m canopy height files for geographic sites identified by NEON four-letter codes. It was created by Ben Weinstein of the University of Florida.

Use Cases

• Train individual tree detection models based on RGB imagery.
• Develop algorithms for canopy height estimation based on LiDAR and canopy height model data.
• Perform species classification or vegetation analysis based on 426-band hyperspectral data.
• Benchmark remote sensing models across multiple geographic sites based on NEON site codes.

Strengths

• Includes up to four complementary data modalities per site: RGB, LiDAR, hyperspectral, and canopy height.
• Annotations for the image tiles are publicly available under version control.
• Data is organized to match the structure of the corresponding evaluation benchmark.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Row count, file formats, and total size are unknown, which may limit suitability assessment.
• Last update date is unknown; freshness unverified.

Provenance

Source

University of Florida

Collection Method

Likely collected as part of the National Ecological Observatory Network (NEON) program.

Geography

Covers multiple geographic sites identified by NEON four-letter codes (e.g., HARV for Harvard Forest).