The canonical location of this page is https://strawlab.org/schemas/flydra/1.3.

flydra v1.3 schema

Description

This is an HDF5 file for saving tracked 3D point data.

Three datasets are in this file:

• /experiment_info - information about the experiment
• /trajectories - estimated 3D point trajectories
• /trajectory_start_times - the first timestamp for each tracked object

experiment_info

The HDF5 description is:

  DATATYPE  H5T_COMPOUND {

     H5T_STRING {

        STRSIZE 32;

        STRPAD H5T_STR_NULLPAD;

        CSET H5T_CSET_ASCII;

        CTYPE H5T_C_S1;

     } "uuid";

  }

The columns are:

• uuid (32 character ASCII string) - an arbitrary identifier unique for each experiment

trajectories

The HDF5 description is:

  DATATYPE  H5T_COMPOUND {

     H5T_STD_U32LE "obj_id";

     H5T_STD_I64LE "framenumber";

     H5T_IEEE_F64LE "x";

     H5T_IEEE_F64LE "y";

     H5T_IEEE_F64LE "z";

     H5T_IEEE_F64LE "covariance_x";

     H5T_IEEE_F64LE "covariance_y";

     H5T_IEEE_F64LE "covariance_z";

  }

The columns are:

• obj_id (uint32) - an arbitrary identifier unique for each trajectory
• framenumber (int64) - the frame for which the estimate is made
• x (float64) and
• y (float64) and
• z (float64) - the (Kalman smoothed) estimated 3D positions.
• covariance_x (float64) and
• covariance_y (float64) and
• covariance_z (float64) - first 3 entries in diagonal of covariance matrix

Additionally, this dataset has the following attributes:

• frames_per_second - the framerate at which the data were acquired.
• smoothed_source - a string equal to kalman_estimates
• smoothed_data_filename - source (filename) of the 3D data
• raw_data_filename - source (filename) of the raw data, including timestamps
• dynamic_model_name - name of the dynamic model used to calculate 3D data
• smoothing_flydra_version - the version number used to produce smoothed 3D data
• recording_flydra_version - the version number used to produce record the data

trajectory_start_times

The HDF5 description is:

  DATATYPE  H5T_COMPOUND {

     H5T_STD_U32LE "obj_id";

     H5T_STD_U64LE "first_timestamp_secs";

     H5T_STD_U64LE "first_timestamp_nsecs";

  }

The columns are:

• obj_id (uint32) - an arbitrary identifier unique for each trajectory
• first_timestamp_secs (uint64) and
• first_timestamp_nsecs (uint64) - the time at which the trajectory started, in seconds and nanoseconds since 1 January 1970. Stored in UTC.

Additionally, this dataset has an attribute "timezone", which contains the timezone in which the data were acquired. Because the timestamps are saved in UTC, this allows displaying them in local time. For the Pacific timezone, this attribute looks like this with h5dump:

  ATTRIBUTE "timezone" {

     DATATYPE  H5T_STRING {

           STRSIZE 10;

           STRPAD H5T_STR_NULLPAD;

           CSET H5T_CSET_ASCII;

           CTYPE H5T_C_S1;

        }

     DATASPACE  SCALAR

     DATA {

     (0): "US/Pacific"

     }

  }

Note about user block in .h5 format

The first bytes of the file are an HDF5 user block that describe the schema of the file. (In other words, it's a descriptor that points to this page.) By putting the first bytes in this file, the freedesktop.org file identification machinery can automatically identify this file.