Analyzing leaf color of 10,950 iNaturalist observations of Red Maple (Acer rubrum)
Using a deep learning based object detector, leaf-detector, leaf positions in each image were detected. From there, the foreground was removed from the detected leaf using GrabCut algorithm. Dominant color of the leaf was calculated using k-means clustering.
Figure 1 (above): All the 2019 data plotted on the same map. The color of each observation marker corresponds to the dominant color of the leaves of the observation.
from leafcolors.leafcolors import LeafColors
detector_output = 'output.txt' # output of leaf-detector batch detection
img_folder = '2015/' # folder of images that darknet detected
my_analysis = LeafColors(detector_output, img_folder)
For the output of leaf-detector, be sure to remove the first few lines of metadata before passing it to LeafColors.