Enhancing Boundary Segmentation for Topological Accuracy with Skeleton-based Methods

Authors: Chuni Liu, Boyuan Ma, Xiaojuan Ban, Yujie Xie, Hao Wang, Weihua Xue, Jingchao Ma, Ke Xu

IJCAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments prove that our method improves topological consistency by up to 7 points in VI compared to 13 state-of-art methods, based on objective and subjective assessments across three different boundary segmentation datasets.
Researcher Affiliation Academia 1Collaborative Innovation Center of Steel Technology, University of Science and Technology Beijing 2Beijing Advanced Innovation Center for Materials Genome Engineering, University of Science and Technology Beijing 3School of Intelligence Science and Technology, University of Science and Technology Beijing 4Shunde Innovation School, University of Science and Technology Beijing 5Institute of Materials Intelligent Technology, Liaoning Academy of Materials 6Institute for Advanced Materials and Technology, University of Science and Technology Beijing 7School of Materials Science and Technology, Liaoning Technical University chuniliu@xs.ustb.edu.cn
Pseudocode No The paper describes the proposed methods in text and uses figures to illustrate concepts (e.g., Fig. 2 (a), (b), (c)), but no explicit pseudocode or algorithm blocks are provided.
Open Source Code Yes The code is available at https://github.com/clovermini/Skea topo.
Open Datasets Yes Our method was evaluated using three publicly available image segmentation datasets from different domains. The first dataset, SNEMI3D [Arganda-Carreras et al., 2013] in biology... The second dataset, Pure Iron Grain(IRON)[Liu et al., 2022]... The third dataset, Massachusetts Roads Dataset (MASS. ROAD)[Mnih, 2013]...
Dataset Splits Yes To assess the model robustness, a threefold cross-validation approach was used for the SNEMI3D and IRON datasets. For the MASS. ROAD Dataset, validation and testing were conducted using its official sets.
Hardware Specification Yes All experiments were performed using an NVIDIA Ge Force RTX 3090 GPU (24GB Memory) and an Intel(R) Xeon(R) Silver 4210R CPU @ 2.40 GHz.
Software Dependencies No The paper mentions 'Py Torch implementation of the UNet model' and 'scikit-image' but does not provide specific version numbers for these or other software dependencies.
Experiment Setup Yes The input size was set to 512 512. During training, the Adam optimizer with default parameters was used, and the Step LR scheduler with a step size of 10 and a decay rate of 0.8 was used to adjust the learning rate. The initial learning rate was set to 1e-4. Each model was trained for 50 epochs with a batch size of 10, and we obtained the best parameter based on early stopping.