Abstract:Contour-based instance segmentation methods include one-stage and multi-stage schemes. These approaches achieve remarkable performance. However, they have to define plenty of points to segment precise masks, which leads to high complexity. We follow this issue and present a single-shot method, called \textbf{VeinMask}, for achieving competitive performance in low design complexity. Concretely, we observe that the leaf locates coarse margins via major veins and grows minor veins to refine twisty parts, which makes it possible to cover any objects accurately. Meanwhile, major and minor veins share the same growth mode, which avoids modeling them separately and ensures model simplicity. Considering the superiorities above, we propose VeinMask to formulate the instance segmentation problem as the simulation of the vein growth process and to predict the major and minor veins in polar coordinates. Besides, centroidness is introduced for instance segmentation tasks to help suppress low-quality instances. Furthermore, a surroundings cross-correlation sensitive (SCCS) module is designed to enhance the feature expression by utilizing the surroundings of each pixel. Additionally, a Residual IoU (R-IoU) loss is formulated to supervise the regression tasks of major and minor veins effectively. Experiments demonstrate that VeinMask performs much better than other contour-based methods in low design complexity. Particularly, our method outperforms existing one-stage contour-based methods on the COCO dataset with almost half the design complexity.

A Leaf Modeling and Multi-Scale Remeshing Method for Visual Computation via Hierarchical Parametric Vein and Margin Representation

View-Dependent Hierarchical Foliage Simplification

Modeling Leaves Based on Real Image

Multiresolution Foliage For Forest Rendering

Efficient Multiresolution of Foliage for Real-Time Rendering

Multi-Resolution Foliage Faithful to the Visual and Botanic Concepts

A Fast Method of Visual Modeling Plant Organs Based on Image

Revealing Hierarchical Structure of Leaf Venations in Plant Science via Label-Efficient Segmentation: Dataset and Method

Grid-based View-Dependent Foliage Simplification

Image-based Plant Modeling by Knowing Leaves from Their Apexes

Multi-Level Tree Branch Modeling and Animation

The Research on Plant's Realistic Modeling Method

Rotation-model Based Simulating Plant Leaf Curling Deformation

Layered Leaf Texturing Using Structure-Guided Model.

Modeling Full-Scale Leaf Venation Networks

Continuous LOD Model of Coniferous Foliage.

Data fusion for a multi-scale model of a wheat leaf surface: a unifying approach using a radial basis function partition of unity method

Leaf Reconstructing Combining L-system and Synthetic Image

Instance Mask Growing on Leaf.

View-Dependent Conifer Lod Models

Geometrical Modeling of leaf Shape Based on Image Processing and Feature Points Detection