imageseg: an R package for deep learning-based image segmentation

Convolutional neural networks (CNNs) and deep learning are powerful and robust tools for ecological applications. CNNs can perform very well in various tasks, especially for visual tasks and image data. Image segmentation (the classification of all pixels in images) is one such task and can for example be used to assess forest vertical and horizontal structure. While such methods have been suggested, widespread adoption in ecological research has been slow, likely due to technical difficulties in implementation of CNNs and lack of toolboxes for ecologists. Here, we present R package imageseg which implements a workflow for general-purpose image segmentation using CNNs and the U-Net architecture in R. The workflow covers data (pre)processing, model training, and predictions. We illustrate the utility of the package with two models for forest structural metrics: tree canopy density and understory vegetation density. We trained the models using large and diverse training data sets from a variety of forest types and biomes, consisting of 3288 canopy images (both canopy cover and hemispherical canopy closure photographs) and 1468 understory vegetation images. Overall classification accuracy of the models was high with a Dice score of 0.91 for the canopy model and 0.89 for the understory vegetation model (assessed with 821 and 367 images, respectively), indicating robustness to variation in input images and good generalization strength across forest types and biomes. The package and its workflow allow simple yet powerful assessments of forest structural metrics using pre-trained models. Furthermore, the package facilitates custom image segmentation with multiple classes and based on color or grayscale images, e.g. in cell biology or for medical images. Our package is free, open source, and available from CRAN. It will enable easier and faster implementation of deep learning-based image segmentation within R for ecological applications and beyond.