UsamaSQ/autosolve-telemetry

AutoSolve Telemetry数据集是一个社区驱动的3D相机跟踪数据集，旨在通过收集匿名跟踪会话来训练自适应学习系统，该系统能够根据素材特征预测最佳跟踪设置（搜索大小、模式大小、运动模型）。数据集包含来自Blender的AutoSolve插件的匿名数值遥测数据，涵盖素材元数据、使用设置、结果、相机内参、运动分析、特征密度、时间序列数据和行为记录。数据集特别强调数据隐私和伦理，仅收集数值遥测数据，不包含图像、视频帧、像素数据、文件路径、项目名称、用户标识符或系统信息。数据集适用于超参数优化、异常检测、运动分类和时间建模等研究任务。

The AutoSolve Telemetry dataset is a community-driven dataset for 3D camera tracking, designed to train an adaptive learning system that predicts optimal tracking settings (Search Size, Pattern Size, Motion Models) based on footage characteristics. It collects anonymized numerical telemetry from tracking sessions via the AutoSolve Blender Addon, including footage metadata, settings used, results, camera intrinsics, motion analysis, feature density, time series data, and behavior records. The dataset emphasizes data privacy and ethics, collecting only numerical telemetry and excluding images, video frames, pixel data, file paths, project names, user identifiers, or system information. It is ideal for research tasks such as hyperparameter optimization, outlier detection, motion classification, and temporal modeling.