nv_cut-dataset

The tool nv_cut-dataset creates a new Dataset from another one, where the new recording is cut in time according to the given start time, stop time, or both start and stop time. Post-process the new recording folder the same way as any normal scan done with the device.

This tool can cut a large dataset and process the two pieces separately. If this large dataset does not use control points and has too much drift, splitting it up also means that you can align the two pieces separately and correct the drift in this way.

Note: Users should use the start-time and stop-time post-processing arguments.

Options

Usage:  
   nv_cut-dataset [-h] --input INPUT --output OUTPUT
                  [--start-time START_TIME] [--stop-time STOP_TIME]
                  [--force]

Options:
  -h, --help            show this help message and exit
  --input INPUT, -i INPUT
                        input recording folder to cut
  --output OUTPUT, -o OUTPUT
                        output folder to write cut dataset to
  --start-time START_TIME
                        start timestamp of cut recording (absolute or relative)
  --stop-time STOP_TIME
                        end timestamp of cut recording (absolute or relative)
  --force, -f           delete destination directory contents if not empty

Usage Examples

The start and stop times are timestamps. They can either be given as absolute stamps since the unix time epoch (01/01/1970 00:00:00) or as relative stamps from the input dataset start time. In both cases, the timestamps must be given in seconds. The tool automatically detects if the given timestamps are absolute or relative.

Work out at which point to cut the dataset by:

• Checking the timestamps of the control points or the info files from datasets_rec/<dataset>/info/.

• Using information from qualitymap.vis.big.png: If processed with a software version 1.6.5 or higher. It includes information about the starting timestamp of the dataset, how many minutes have elapsed since the start of the dataset, and the capture location index.

Here is an example section of the quality map from a larger dataset processed with this feature:

Cut the given dataset to create a new dataset with start-time 1481036112.0 and stop-time 1481036200.0:

nv_cut-dataset --input=/srv/data/datasets_rec/2016-12-12_16.03.15 --start-time=1481036112.0 \
        --stop-time=1481036200.0 --output=/srv/data/datasets_rec/2016-12-12_16.03.15_cut

Cut the given dataset to create a new dataset starting 20 seconds later than the input dataset start time and stopping 60 seconds later than the input dataset start time:

nv_cut-dataset --input=/srv/data/datasets_rec/2016-12-12_16.03.15 --start-time=20.0 \
        --stop-time=60.0 --output=/srv/data/datasets_rec/2016-12-12_16.03.15_cut_start

The output dataset looks just like any other dataset; as if it had been captured only between the given start and stop times. The only difference is a new file named cut_info, which contains the dataset's original start and stop times. The quality map is also evidence that the dataset was cut: because nv_cut-dataset does not re-run SLAM, the quality map shows the original dataset.

Cut the given dataset to create a new dataset with start-time 1481036112.0 and a stop time equal to the old dataset stop time:

nv_cut-dataset --input=/srv/data/datasets_rec/2016-12-12_16.03.15 --start-time=1481036112.0 \
        --output=/srv/data/datasets_rec/2016-12-12_16.03.15_cut_start

Cut the given dataset to create a new dataset with the old dataset start time and stop-time 1481036200.0:

nv_cut-dataset --input=/srv/data/datasets_rec/2016-12-12_16.03.15 \
        --stop-time=1481036200.0 --output=/srv/data/datasets_rec/2016-12-12_16.03.15_cut_end