In recent years latent SVM and discriminative parts models for object detection are one of the most talked about and evaluated codes. The reference site for this work is http://www.cs.berkeley.edu/~rbg/latent/
They have a library for object detection that runs on Matlab in linux and Mac OS. Here are the steps I followed to set it up. There is bug in one of the files and it took me sometime to figure it out.
Download the library from http://www.cs.berkeley.edu/~rbg/latent/ and unzip the contents
The Readme file can be used very easily to run the files. Use compile.m to compile the mex files.
While doing this step, you will face a error with mex file saying -o option is not available (I have informed the author about this error; so if you dont get this error, then great).
Goto the lines:
eval([mexcmd ' gdetect/fconvsse.cc -o fconv']);eval([mexcmd ' gdetect/fconv_var_dim_MT.cc -output fconv_var_dim']);
Change -o to -output
Thats all. It should run.