MiCT: Mixed 3D/2D Convolutional Tube for Human Action Recognition

Yizhou Zhou, Xiaoyan Sun, Zheng-Jun Zha, Wenjun Zeng

URL

Proceedings: CVPR

Year: 2018

Peer Reviewed : Yes

Results: 
HMDB /HMDB Eval: 63.8% (MiCT-Net)
UCF101 /UCF101 Eval: 88.9% (MiCT-Net)
HMDB /HMDB Eval: 70.5% (Two-stream MiCT-Net)
UCF101 /UCF101 Eval: 94.7% (Two-stream MiCT-Net)