erlingdevold / CRIMAC-classifiers-bottom

Bottom detection algorithms

CRIMAC-classifiers-bottom

This repository contains code for bottom detection in the CRIMAC project.

The bottom detection works on zarr files created by CRIMAC-preprocessing.

https://hub.docker.com/r/crimac/bottomdetection

Running using Docker

Mounted directories:

1. /in_dir - Input directory containing zarr data.
2. /out_dir - Output directory where the annotation file will be written.
3. /work_dir - Directory with LSSS work files. Needed only when ALGORITHM is set to work_files.

Options as environment variables:

1. INPUT_NAME - Name of the zarr file in in_dir.

2. OUTPUT_NAME - Name of the annotation file in out_dir. The output format is given by the file name suffix.

   • Pandas DataFrame:
     • .csv
     • .html
     • .parquet
   • Xarray Dataset:
     • .nc
     • .zarr

3. ALGORITHM - Optional. The bottom detection algorithm to use:

   • angles - For data with sloping bottom. Used by the combined algorithm.
   • combined - A combination of the edge and angles algorithms.
   • constant - A very fast algorithm for testing and debugging.
   • edge - For data with flat bottom. Used by the combined algorithm.
   • simple - Backstepping from maximum s_v.
   • work_files - Uses the lowest layer boundary in LSSS work files as the detected bottom. The /work_dir directory must be mounted.

   See more details in a separate document on the bottom detection algorithms.

4. Algorithm parameters. Optional.

   • PARAMETER_minimum_range [m] - The minimum range of the detected bottom.
   • PARAMETER_offset [m] - Additional offset to the bottom after backstepping.
   • PARAMETER_threshold_log_sv [dB] - The minimum S_v value for detecting bottom.

Example:

docker run -it --name bottomdetection \
  -v /home/user/data:/in_dir \
  -v /home/user/output:/out_dir  \
  --env INPUT_NAME=dataset.zarr \
  --env OUTPUT_NAME=bottom.parquet \
  --env ALGORITHM=simple \
  --env PARAMETER_offset=0.5 \
  crimac/bottomdetection