LATENT TRAINING OF CASCADE DETECTOR OF CONVEX CLASSIFIERS

  • The programs run on Linux 64 bit machines (they use compiled mex files).
  • Download LatentTrain.rar files and uncompress them in an appropriate directory.
  • Download PASCAL VOC 2007 datasets and development kit from VOC2007
  • You must set the appropriate directory names in “globals_latent_training.m” file.
  • Run “latent_training_detector_cascade.m”. It will create the detector model of the object class you choose. You can set the number of roots and the names of the object classes and other parameters at the beginning of the file.
  • You can determine the nonlinear classifier kernel parameters by using “nonlinear_HS_CV.m” for kernel hypersphere classifier and “nonlinear_SVM_CV.m” for kernel SVM classifiers. You need to determine these parameters before Phase 3 of the algorithm.
  • If you want to use our created object class models, some of them are in pretrained_models folder.

If you find a bug, please send an email to hakan.cevikalp@gmail.com