RunningFelzTraining < LCI

Tags: view all tags

Follow following steps to run training for Felzenszwalb detector-
   * Make sure you have right directory structure read in _training_data_ node(described [[AnnotatingData][here]])
   * Add a training file to ImageSets with name category_train.txt (e.g., toy_Stegosaurus_train.txt)
   * Add list of xml files(without extension) as positive examples.
   * Add list of jpg images(without extension) for negative examples.
   * *pyUtil* package has a node _list_files_ to make above two steps easier(type $rosrun pyUtils list_files.py --help )
   * Modify(or create a new)launch file _dpm_train.launch_ to enter category you'd like to train for.(make sure you have a corresponding train file available, see second step)
   * Run $ roslaunch DPM dpm_train.launch

-- Main.AnkurGupta - 12 Nov 2009

Raw edit | More topic actions

Topic revision: r1 - 2009-11-12 - AnkurGupta