Remote sensing of forested landscapes can transform the speed, scale, and cost of forest research. Data acquisition currently outpaces the ability to detect individual trees in high resolution imagery. Here we introduce a new python package, DeepForest, the first open source implementation of a deep learning neural network for RGB crown detection. Deep learning has made enormous strides in a range of computer vision tasks but requires significant amounts of training data. By including a prebuilt model, DeepForest will simplify the process of using and retraining deep learning models for a range of forests, sensors, and spatial resolutions.
Results/Conclusions
We illustrate the performance of DeepForest on an evaluation dataset from the National Ecological Observation Network that covers a range of forest types. We then applied DeepForest to the entire NEON catalog of RGB data encompassing fifty geographic sites and tens of thousands of remote sensing tiles. Using co-registered LiDAR data, we extract information on tree height and size for over 50 million trees across the continental United States.