SSCV is an open source MATLAB implementation of the super sparse coding vector (SSCV) for human activity recognition. This package illustrates how to compute SSCV on the YouTube dataset.
The code depends on SPAMS library which is enclosed in the package. If it is not compatible with your platform, please download and setup accordingly.
Please extract STIP of the YouTube dataset and put the feature files (*.stip) under "./Feats/STIP/" according to the structures and names in this dataset.
If you find any bug or develop some extensions, please feel free to drop me a line.
Please cite our paper if you use the code:
X. Yang and Y. Tian. Action Recognition Using Super Sparse Coding Vector with Spatio-Temporal Awareness. European Conference on Computer Vision (ECCV), 2014. [PDF]
Copyright (C) 2015 Xiaodong Yang
Distibution code version 1.0 - 01/24/2015. The code is for research purpose only.