Actionness-assisted Recognition of Actions

Oct, 2015

Abstract

We elicit from a fundamental definition of action low-level attributes that can reveal agency and intentionality. These descriptors are mainly trajectory-based, measuring sudden changes, temporal synchrony, and repetitiveness. The actionness map can be used to localize actions in a way that is generic across action and agent types. Furthermore, it also groups interacting regions into a useful unit of analysis, which is crucial for recognition of actions involving interactions. We then implement an actionness-driven pooling scheme to improve action recognition performance. Experimental results on various datasets show the advantages of our method on action detection and action recognition comparing with other state-of-the-art methods.

[ Paper ], [ BibTex ]


Method

We define a bottom-up actionness model and develop a computational model of the actionness value. Our actionness attributes consist of

Then we group temporal superpixels into spatial-temporal regions according the actionness values. Action pooling operation is carried on these regions.


Results

Action detection performance (mAP reported)

Method Our method L-CORF
DPM
UCF-Sports 66.81 60.8 54.9
HOHA 70.16 68.5 60.8

Action recognition performance (average accuracy reported)

Dataset SSBD HDMB51
UCF50
Actionness-pooling + FV + linear SVM 76 60.41 92.48


Code

[BoW frameworks]


Demos


How to cite

If you find our work useful, please cite it. The bibtex entry is provided below for your convenience.

[ Paper ], [ BibTex ]


			@inproceedings{YeLuo_ICCV_2015,
				author    = {Ye Luo and
				Loong-Fah Cheong and
				An Tran},
				title     = {Actionness-assisted Recognition of Actions},
				booktitle = {The IEEE International Conference on Computer Vision (ICCV)},
				year      = {2015},
			}
		


Contact

For any questions regarding the work or the implementation, contact the authors at