Question

I spent days to search about this topic and I found some good articles, but all them discuss about theory, and I have to implement it using MATLAB. Specified object are picked from Caltec 101 dataset. I know to reach this goal there are many methods, using objects geometric features, template matching, and other methods. Is there any tutorial to implement one of this methods in MATLAB?

*I can choose another dataset, it’s not forced to use Caltec 101

Caltec 101:

Another dataset

Answer 1

If you want something that works almost directly “off-the-shelf” and that is mostly coded in MATLAB, you should try the framework developed for “Discriminatively Trained Deformable Part Models” and subsequent papers in PAMI 2010 and CVPR 10. You can find all the code for this framework on a site hosted by the authors.

Even though the title of the first paper may suggest that the system is designed to deal with objects with deformable parts only, the system works on a wide variety of object categories, rigid or with deformable parts.

This system also won the Lifetime Achievement Prize from the PASCAL Visual Object Classes Challenge Workshop 2010, in recognition for (I quote):

“First submitted to the PASCAL VOC challenge in 2007, this detector
has now become a core component of many classification, segmentation,
person layout and action classification submissions. We honour the
contribution made to the community by the innovation and success of
the method and its free distribution.”