Procedural Generation of Videos to Train Deep Action Recognition Networks

César Roberto de Souza, Adrien Gaidon, Yohann Cabon, Antonio Manuel López Peña

URL

Proceedings: IEEE Conference on Computer Vision and Pattern Recognition (CVPR)

Year: 2017

Peer Reviewed : No

Results: 
HMDB /HMDB Eval: 69.5% (Leveraging our synthetic dataset and multi-task models, we increase the performance from 66.6 to 69.5)
UCF101 /UCF101 Eval: 94.2% (Leveraging our synthetic dataset and multi-task models, we increase the performance from 93.6 to 94.2)