Manipulated Object Proposal: A Discriminative Object Extraction and Feature Fusion Framework for First-Person Daily Activity Recognition

no code implementations • 2 Sep 2015 • Changzhi Luo, Bingbing Ni, Jun Yuan, Jian-Feng Wang, Shuicheng Yan, Meng Wang

This scheme leverages motion cues such as motion boundary and motion magnitude (in contrast, camera motion is usually considered as "noise" for most previous methods) to generate a more compact and discriminative set of object proposals, which are more closely related to the objects which are being manipulated.

Action Recognition Object +2