Tools

The eblearn tools help to create datasets, train models and run them.
Their code is located in eblearn/tools/tools/src and binaries are built into eblearn/bin, and installed on the system with 'make install'.
They can all be compiled by calling 'make tool', or simply 'make'.
Most tools will show a brief help when called without arguments.

Dataset tools

  • dscompile: preprocess and compile images into a dataset ready for training.
  • dssplit: split a dataset in two (e.g. validation and training).
  • dsmerge: merge two datasets into one (e.g. positive and negative datasets).
  • dsdisplay: show samples and labels contained in a dataset.

Training tools

  • train: train a model given a configuration file.
  • maketable: create a table matrix for convolutional connections.

Detection tools

  • detect: find bounding boxes around objects in images given a trained model (multi-threaded version).
  • track: find and track objects in video inputs given a trained model.
  • stdetect: single-threaded version of detect.
  • mpidetect: cluster multi-threaded version of detect, using MPI.

Jobs tools

  • metarun: run multiple jobs in parallel or in sequence, given all possible configurations in a configuration file with multi-values variables. metarun analyzes outputs and sends emails with reports and curves at different iterations. It can also be used with clusters (using MPI).
  • metaparse: parse the output files and directories of metarun, analyze results and send reports and curves.

Display tools

  • matshow: display images of any type and .mat matrix types. matshow can show multiple images at the same time. It can also display weights of a stored model.

Capture tools

  • capture: capture images from camera for training.