Organizing your data

Data directory structure

Note

The Lightning Pose CLI historically supported flexible data organization, as written in this document. The Lightning Pose App requires a more constrained directory structure.

See the Singleview Data Directory Structure for the detailed reference of the expected directory structure that is fully app-compatible.

Lightning Pose assumes the following project directory structure, as in the example dataset here.

/path/to/project/
  ├── project.yaml
  ├── <LABELED_DATA_DIR>/
  ├── <VIDEO_DIR>/
  └── <YOUR_LABELED_FRAMES>.csv

  • project.yaml: contains project-level metadata. See Project Yaml File Format.

  • <YOUR_LABELED_FRAMES>.csv: a table with keypoint labels (rows: frames; columns: keypoints). Note that this file can take any name, and needs to be specified in the config file under data.csv_file. For more details on the format, see Label CSV File Format.

  • <LABELED_DATA_DIR>/: contains images that correspond to the labels, and can include subdirectories. The directory name, any subdirectory names, and image names are all flexible, as long as they are consistent with the first column of <YOUR_LABELED_FRAMES>.csv.

  • <VIDEO_DIR>/: when training semi-supervised models, the videos in this directory will be used for computing the unsupervised losses. This directory can take any name, and needs to be specified in the config file under data.video_dir. Notes that videos must be mp4 files that use the h.264 codec; see more information in the FAQs.