DLTcode

  Robust Non-negative Dictionary Learning for Visual Tracking The provided codes could be either embedded into the benchmark framework of paper Online Object Tracking: A Benchmark (CVPR2013) (You can find details here: http://visual-tracking.net/) or run on individual sequence. To run the benchmark, just put the entire folder into the /trackers folder in the benchmark code base, and modify the configTrackers.m in util folder. DLT gets an AUC of 0.436, which ranks 5th among 26 in the benchmark by 19/03/2014. We don t tune parameters for single sequence in this case, all the parameters are stored in trackparam_DLT.m. To run on individual video, you need to modify the dataPath and title in run_individual.m. If you run MATLAB version after 2012, and have a CUDA compatible GPU installed, you may enjoy the fast computation speed by GPU, just set useGPU to true in trackparam_DLT.m and run_individual.m!