DeepD3 Datasets

In this repository, you will find: The DeepD3 benchmark dataset - a TIFF stack with an extent of 135.7 by 34.4 by 35.5 µm from the rat hippocampal CA1 region (94 x 94 x 500 nm voxel size) The dendrite tracings of the DeepD3 benchmark dataset by three human annotators U, V and W in SWC file format, traced with neutube The pixel-precise spine annotations of the DeepD3 benchmark dataset by three human annotators as *.mask files (compatible with pipra, but easy to open with any HDF5 reader) The human annotations and clusters for the DeepD3 benchmark dataset as CSV file A Jupyter notebook that shows how we assigned the human annotations to individual clusters that pertain to individual dendritic spines The DeepD3 training data as d3set file The DeepD3 validation data as d3set file (see paper for further information, basically HDF5 files)本仓库包含以下内容：DeepD3基准数据集（DeepD3 benchmark dataset）—— 取自大鼠海马CA1区的TIFF图像栈（TIFF stack），视场范围为135.7×34.4×35.5 µm，体素尺寸为94×94×500 nm；由三位人工标注者U、V、W借助neutube软件追踪得到的DeepD3基准数据集树突追踪文件，格式为SWC文件（SWC file）；三位人工标注者为该基准数据集制作的像素级精确树突棘标注文件，格式为*.mask文件（兼容pipra工具，可通过任意HDF5阅读器轻松打开）；以CSV文件格式存储的DeepD3基准数据集人工标注及聚类结果；一份Jupyter Notebook，用于演示如何将人工标注分配至对应单个树突棘的聚类组别；以d3set文件格式存储的DeepD3训练数据集；以d3set文件格式存储的DeepD3验证数据集（详细信息参见原论文，其本质为HDF5文件）