AxonEM

AxonEM 数据集由分别来自人和小鼠皮层的两个 30x30x30 um^3 EM 图像卷组成。它用于大脑皮层区域的 3D 轴突实例分割。作者校对了 18,000 多个轴突实例，以提供密集的 3D 轴突实例分割，从而能够对轴突重建方法进行大规模评估。此外，作者还为每个数据卷密集标注了 9 个用于训练的 ground truth 子卷。

The AxonEM dataset consists of two 30×30×30 μm³ EM image volumes sourced from the cerebral cortices of humans and mice, respectively. It is designed for 3D axon instance segmentation tasks in cerebral cortical regions. The authors have manually proofread over 18,000 axon instances to generate dense 3D axon instance segmentation annotations, enabling large-scale evaluation of axon reconstruction methods. Additionally, the authors have densely annotated 9 ground-truth sub-volumes for model training from each of the two volumes.

搜集汇总

背景与挑战

背景概述

AxonEM数据集提供人和小鼠皮层的3D电子显微镜图像数据，专门用于大脑皮层区域的轴突实例分割。该数据集包含超过18,000个校正的轴突实例标注以及9个训练子卷，旨在支持轴突重建方法的大规模评估。