"data"

"This document specifies the requirements and procedures for uploading raw squat motion data to support model training, validation, and testing for motion recognition, biomechanical analysis, and posture evaluation tasks. All data must be uploaded in unprocessed, original formats including time-series kinematic data, joint angle sequences, IMU sensor readings, 3D keypoint coordinates, force plate measurements, and original video frames. Please retain complete timestamps, sampling rates, sensor IDs, subject information, and movement labels to ensure data traceability and model generalization.Files should be organized in standardized formats such as CSV, JSON, or Parquet, with clear naming conventions indicating subject ID, movement type, trial number, and collection date. Missing values, outliers, and duplicate entries must not be pre\u2011removed; raw noise and natural variations are critical for robust model training. Each upload must include a metadata file documenting collection equipment, environment, sampling frequency, and annotation rules.Strictly follow platform path specifications and file size limits; compress large datasets without altering internal structure. Label consistency and data integrity directly affect model accuracy, convergence speed, and reliability of squat quality classification and injury risk prediction. Incomplete or non\u2011standard submissions will be returned for revision to guarantee high\u2011quality data input for machine learning pipelines."