Urban Tree Detection Data

This repository provides a dataset for training and evaluating tree detectors in urban environments with aerial imagery. The dataset includes:

• 256x256 crops of 60 cm aerial imagery from the 2016, 2018, and 2020 NAIP across eight cities in California
• A total of 1,651 images and 95,972 annotated trees
• Point annotations for all trees visible in the imagery
• Train/val/test splits to replicate or compare against the results in our paper

Data description

The dataset covers eight cities and six climate zones in California across three years. The following table provides a summary. The three right-most columns give the number of annotated trees in each year.

City	Climate Zone	Number of Crops	2016	2018	2020
Bishop	Interior West	10	-	-	682
Chico	Inland Valleys	99	-	8,187	8,164
Claremont	Inland Empire	92	4,858	4,880	4,678
Eureka	Northern California Coast	21	-	-	2,134
Long Beach	Southern California Coast	100	6,470	6,403	5,845
Palm Springs	Southwest Desert	100	4,433	4,707	4,109
Riverside	Inland Empire	90	5,015	4,400	4,087
Santa Monica	Southern California Coast	92	5,824	5,830	5,841

The bands in the imagery are as follows:

Band	Description
0	Red
1	Green
2	Blue
3	Near-IR

Data organization

• Images are stored in the images directory as TIFF files.
• Each image has an associated CSV file in the csv directory containing tree locations in 2D pixel coordinates.
• Each image has an associated GeoJSON file in the json directory containing geo-referenced tree locations. Coordinates are stored in the local UTM zone.
• A missing .csv or .json file means that there are no trees in the image.

The files train.txt, val.txt, and test.txt specify the splits using all of the data. The files train_socal.txt, val_socal.txt, and test_socal.txt specify the splits using the Southern California 2020 subset of the data (only 2020 data from Claremont, Long Beach, Palm Springs, Riverside, and Santa Monica).

Citation

NAIP on AWS was accessed on January 28, 2022 from https://registry.opendata.aws/naip.

If you use this data, please cite our paper:

J. Ventura, C. Pawlak, M. Honsberger, C. Gonsalves, J. Rice, N.L.R. Love, S. Han, V. Nguyen, K. Sugano, J. Doremus, G.A. Fricker, J. Yost, and M. Ritter (2024). Individual Tree Detection in Large-Scale Urban Environments using High-Resolution Multispectral Imagery. International Journal of Applied Earth Observation and Geoinformation, 130, 103848.

Acknowledgments

This project was funded by CAL FIRE (award number: 8GB18415) the US Forest Service (award number: 21-CS-11052021-201), and an incubation grant from the Data Science Strategic Research Initiative at California Polytechnic State University.

License

This work is licensed under a Creative Commons Attribution 4.0 International License.