Automatic detection and segmentation of overlapping leaves in dense foliage can be a difficult task, particularly for leaves with strong textures and high occlusions. We present Dense-Leaves, an image dataset with ground truth segmentation labels that can be used to train and quantify algorithms for leaf segmentation in the wild. It consists of 108 images at resolution 1024 x 768, each of dense foliage from trees, vines and bushes on or near the Michigan State University campus. Roughly 20 leaves per image have been manually segmented. As an initial augmentation step, each image and segmentation has been rotated 6 times by 15 degrees.
 |
 |
Download the Dense-Leaves dataest here (827 MB, May 2018)
More details are in the paper A Pyramid CNN for Dense-Leaves Segmentation
Download code for running and evaluating the Pyramid CNN trained for leaf boundary detection, plus segmentation code
Code for training will be available soon
