DLT (deeplearningtracker) Learning and Code Understanding (1)

Continuous update ...

The DLT is presented by Naiyan Wang and Dit-yan yeung of Hkust, and the articles and codes are shown in http://winsty.net/dlt.html

Related papers are learning a deep Compact Image representation for Visual Tracking (NIPS2013 ')

The paper itself does not have too many bright spots, can say basically did not speak what.

The general idea is to use Sdae (stacked denoising Autoencoder) to train a lot of data offline, the purpose of off-line training is to extract features. And then use it for online tracking. Tracking is done using particle filtering.

Let's start with the research code.

RUN_INDIVIDUAL.M just provides some image information and parameters, the core function is implemented in RUN_DLT.M.

DLT (deeplearningtracker) Learning and Code Understanding (1)