R. Wang, X.-J. Wu, and J. Kittler, "SymNet: A Simple Symmetric Positive Definite Manifold Deep Learning Method for Image Set Classification, " in IEEE Transactions on Neural Networks and Learning Systems, 2020.
(1) deepmain.m is the main file, which implements the structure of SymNet-v1; (2) computeCov.m is constructed to compute the SPD matrices for the training and test image sets (video clips); (3) fun_SymNet_Train.m is applied to implement the KDA algorithm.
(1) place the four .mat files in the folder of SymNet to the folder of SymNet-v1; (2) run deepmain.m.
Matlab R2019a software
(1) FPHA_train_seq.mat and FPHA_train_label.mat are the training samples and the corresponding label information, respectively; (2) FPHA_val_seq.mat and FPHA_val_label.mat are the test samples and the corresponding label information, respectively; (3) This dataset is provided by \cite{FPHA}. Please kindly refer to it. @inproceedings{FPHA, title={First-person hand action benchmark with rgb-d videos and 3d hand pose annotations}, author={Garcia-Hernando, Guillermo and Yuan, Shanxin and Baek, Seungryul and Kim, Tae-Kyun}, booktitle={Proceedings of the IEEE conference on computer vision and pattern recognition}, pages={409--419}, year={2018} }
if you want to run SymNet-v2, you should place the four .mat files to SymNet-v2 folder,firstly. Then, run deepmain_v2.m. After a few seconds, the classification accuracy will be output.
(1) SymNet-v1: 81.04% (2) SymNet-v2: 82.96%