TY - GEN
T1 - Action anticipation with RBF kernelized feature mapping RNN
AU - Shi, Yuge
AU - Fernando, Basura
AU - Hartley, Richard
N1 - Publisher Copyright:
© Springer Nature Switzerland AG 2018.
PY - 2018
Y1 - 2018
N2 - We introduce a novel Recurrent Neural Network-based algorithm for future video feature generation and action anticipation called feature mapping RNN. Our novel RNN architecture builds upon three effective principles of machine learning, namely parameter sharing, Radial Basis Function kernels and adversarial training. Using only some of the earliest frames of a video, the feature mapping RNN is able to generate future features with a fraction of the parameters needed in traditional RNN. By feeding these future features into a simple multilayer perceptron facilitated with an RBF kernel layer, we are able to accurately predict the action in the video. In our experiments, we obtain 18% improvement on JHMDB-21 dataset, 6% on UCF101-24 and 13% improvement on UT-Interaction datasets over prior state-of-the-art for action anticipation.
AB - We introduce a novel Recurrent Neural Network-based algorithm for future video feature generation and action anticipation called feature mapping RNN. Our novel RNN architecture builds upon three effective principles of machine learning, namely parameter sharing, Radial Basis Function kernels and adversarial training. Using only some of the earliest frames of a video, the feature mapping RNN is able to generate future features with a fraction of the parameters needed in traditional RNN. By feeding these future features into a simple multilayer perceptron facilitated with an RBF kernel layer, we are able to accurately predict the action in the video. In our experiments, we obtain 18% improvement on JHMDB-21 dataset, 6% on UCF101-24 and 13% improvement on UT-Interaction datasets over prior state-of-the-art for action anticipation.
KW - Adversarial training
KW - Human action prediction
KW - Radial Basis Function kernel
KW - novel Recurrent Neural Network
UR - http://www.scopus.com/inward/record.url?scp=85055090467&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-01249-6_19
DO - 10.1007/978-3-030-01249-6_19
M3 - Conference contribution
SN - 9783030012489
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 305
EP - 322
BT - Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings
A2 - Hebert, Martial
A2 - Ferrari, Vittorio
A2 - Sminchisescu, Cristian
A2 - Weiss, Yair
PB - Springer Verlag
T2 - 15th European Conference on Computer Vision, ECCV 2018
Y2 - 8 September 2018 through 14 September 2018
ER -