Here we provide the code of actionness estimation with hybrid fully convolutional networks from the following paper:
Actionness Estimation Using Hybrid Fully Convolutional Netoworks
Limin Wang, Yu Qiao, Xiaou Tang, and Luc Van Gool, in CVPR, 2016
- Jul 22, 2016
- Initilaize repo of actionness estimation.
- demo_a_fcn.m: an example showing actionness estimation with A-FCN.
- demo_m_fcn.m: an example showing actionness estimation with M-FCN.
- For optical flow extraction, we use TVL1 Optical Flow
You need download our dense flow code and compile it by yourself. Dense Flow
- Actionness estimation models (A-FCN) on the dataset of Stanford 40:
http://mmlab.siat.ac.cn/actionness/stanford40_actionness_a-fcn.caffemodel - Actionness estimation models (A-FCN and M-FCN) on the dataset of UCF Sports:
http://mmlab.siat.ac.cn/actionness/ucf_sports_actionness_a-fcn.caffemodel http://mmlab.siat.ac.cn/actionness/ucf_sports_actionness_m-fcn.caffemodel - Actionness estimation models (A-FCN and M-FCN) on the dataset of JHMDB:
http://mmlab.siat.ac.cn/actionness/jhmdb_split1_actionness_a-fcn.caffemodel http://mmlab.siat.ac.cn/actionness/jhmdb_split1_actionness_m-fcn.caffemodel
http://mmlab.siat.ac.cn/actionness/jhmdb_split2_actionness_a-fcn.caffemodel http://mmlab.siat.ac.cn/actionness/jhmdb_split2_actionness_m-fcn.caffemodel
http://mmlab.siat.ac.cn/actionness/jhmdb_split3_actionness_a-fcn.caffemodel http://mmlab.siat.ac.cn/actionness/jhmdb_split3_actionness_m-fcn.caffemodel
Contact