A Pyramid CNN for Dense-Leaves Segmentation

TL;DR

This paper introduces Dense-Leaves, a new dataset for leaf segmentation, and proposes a pyramid CNN with multi-scale predictions to improve dense foliage leaf segmentation, achieving promising results.

Contribution

The paper presents a new dataset and a novel pyramid CNN architecture for dense-leaves segmentation, enhancing detection and discrimination of overlapping leaves.

Findings

Effective segmentation of dense foliage leaves.

High accuracy in leaf boundary detection.

Robustness to occlusions and textures.

Abstract

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. We also propose a pyramid convolutional neural network with multi-scale predictions that detects and discriminates leaf boundaries from interior textures. Using these detected boundaries, closed-contour boundaries around individual leaves are estimated with a watershed-based algorithm. The result is an instance segmenter for dense leaves. Promising segmentation results for leaves in dense foliage are obtained.