Learner

diffeo-learner program

This program is used to learn a diffeomorphism model.

The usage is as follows:

$ diffeo-learner --input <input>.processed.bag
                 --path <path>
                 --name <name of model learned>
                 --size [W,H] 
                 --area [rx,ry]

Input

• <input>.processed.bag: ROS bag produced by the Preprocessor. These contains the streams:

  /Y0 -> Image
  /Y1 -> Image /U0 -> int[2]

Y0, U0 and the corresponding Y1 all have the same timestamps.

• <path> and ``>`: path to output file and name of output file.

• Learning parameters:

  • size: size of the image, may be determined from the logs instead of passed, TO WRITE
  • area: search area size in pixels

Output

The output is in a couple of files:

• <path>/<name>.discdds.pickle -- Serialized object of class DiffeoSystem.
• <path>/<name>.discdds.yaml -- Corresponding configuration for dp.