ExoFILT: Transfer learning for robust and accelerated analysis of exocytosis single-particle tracking data

This dataset provides microscopy data to test the software (ImageJ scripts + Jupyter Notebooks) available in the GitHub repository associated with the ExoFILT publication. Three compressed zip files are provided. Two of them contain four .tif movies corresponding to raw imaging data from the datasets RD1 and RD2. The third zip file contains 53 movies from the case study on exocyst-Sec1. In the ExoFILT study: RD1 was used to assess inter-annotator agreement and evaluate neural network performance. RD2 was used for training the ExoFILT model. Additionally, two .csv files containing manual annotations are provided: Annotations from multiple annotators (Ann1-Ann5) on a filtered dataset (FD1) resulting from RD1. Annotations from a single annotator (Ann1) on a filtered dataset (FD2) resulting from RD2. These .csv files can be used together with the annotation GUI available in the GitHub repository to visualize examples of bona fide and ambiguous exocytic events. Overall, these files allow users to reproduce key steps of the analysis pipeline and test the preprocessing, tracking, and inference workflows described in the repository.本数据集提供显微成像数据，用于测试与ExoFILT论文关联的GitHub仓库中发布的软件（含ImageJ脚本与Jupyter Notebook）。 本次共提供三个压缩zip文件。其中两个压缩包各包含4段.tif格式影像，分别对应数据集RD1与RD2的原始成像数据；第三个压缩包包含53段来自胞外囊泡-Sec1案例研究的影像。 在ExoFILT研究中： RD1 用于评估标注者间一致性，并评测神经网络性能； RD2 用于训练ExoFILT模型。 此外，本次还提供两份包含人工标注结果的.csv文件： 1. 多位标注者（Ann1至Ann5）对RD1生成的过滤后数据集FD1的标注结果； 2. 单一位标注者（Ann1）对RD2生成的过滤后数据集FD2的标注结果。 上述.csv文件可与GitHub仓库中提供的标注图形用户界面（GUI）配合使用，以可视化真实胞吐事件与模棱两可的胞吐事件示例。 综上，这些文件可帮助用户复现分析流程的关键步骤，并测试仓库中描述的预处理、追踪与推理工作流。