Featureless:  Bypassing feature extraction in action categorization

Silvia Pintea; Pascal Mettes; Jan van Gemert; AWM Smeulders

doi:10.1109/ICIP.2016.7532346

Featureless: Bypassing feature extraction in action categorization

Silvia Pintea, Pascal Mettes, Jan van Gemert, AWM Smeulders

Pattern Recognition and Bioinformatics

Research output: Chapter in Book/Conference proceedings/Edited volume › Conference contribution › Scientific › peer-review

2 Citations (Scopus)

17 Downloads (Pure)

Abstract

This method introduces an efficient manner of learning action categories without the need of feature estimation. The approach starts from low-level values, in a similar style to the successful CNN methods. However, rather than extracting general image features, we learn to predict specific video representations from raw video data. The benefit of such an approach is that at the same computational expense it can predict 2D video representations as well as 3D ones, based on motion. The proposed model relies on discriminative Wald-boost, which we enhance to a multiclass formulation for the purpose of learning video representations. The suitability of the proposed approach as well as its time efficiency are tested on the UCF11 action recognition dataset.

Original language	English
Title of host publication	2016 IEEE International Conference on Image Processing (ICIP)
Subtitle of host publication	Proceedings
Place of Publication	Piscataway
Publisher	IEEE
Pages	196-200
Number of pages	5
ISBN (Electronic)	978-1-4673-9961-6
ISBN (Print)	978-1-4673-9962-3
DOIs	https://doi.org/10.1109/ICIP.2016.7532346
Publication status	Published - 2016
Event	2016 IEEE International Conference on Image Processing (ICIP) - Phoenix, United States Duration: 25 Sept 2016 → 28 Sept 2016

Conference

Conference	2016 IEEE International Conference on Image Processing (ICIP)
Country/Territory	United States
City	Phoenix
Period	25/09/16 → 28/09/16

Keywords

Multiclass Waldboost
video representations
action recognition
feature learning

Access to Document

10.1109/ICIP.2016.7532346

1803.06962Accepted author manuscript, 514 KB

Cite this

@inproceedings{c8b86420aece4835a54ad97457dc8841,

title = "Featureless: Bypassing feature extraction in action categorization",

abstract = "This method introduces an efficient manner of learning action categories without the need of feature estimation. The approach starts from low-level values, in a similar style to the successful CNN methods. However, rather than extracting general image features, we learn to predict specific video representations from raw video data. The benefit of such an approach is that at the same computational expense it can predict 2D video representations as well as 3D ones, based on motion. The proposed model relies on discriminative Wald-boost, which we enhance to a multiclass formulation for the purpose of learning video representations. The suitability of the proposed approach as well as its time efficiency are tested on the UCF11 action recognition dataset.",

keywords = "Multiclass Waldboost, video representations, action recognition, feature learning",

author = "Silvia Pintea and Pascal Mettes and {van Gemert}, Jan and AWM Smeulders",

year = "2016",

doi = "10.1109/ICIP.2016.7532346",

language = "English",

isbn = "978-1-4673-9962-3 ",

pages = "196--200",

booktitle = "2016 IEEE International Conference on Image Processing (ICIP)",

publisher = "IEEE",

address = "United States",

note = "2016 IEEE International Conference on Image Processing (ICIP) ; Conference date: 25-09-2016 Through 28-09-2016",

}

Pintea, S, Mettes, P, van Gemert, J & Smeulders, AWM 2016, Featureless: Bypassing feature extraction in action categorization. in 2016 IEEE International Conference on Image Processing (ICIP): Proceedings. IEEE, Piscataway, pp. 196-200, 2016 IEEE International Conference on Image Processing (ICIP), Phoenix, United States, 25/09/16. https://doi.org/10.1109/ICIP.2016.7532346

TY - GEN

T1 - Featureless

T2 - 2016 IEEE International Conference on Image Processing (ICIP)

AU - Pintea, Silvia

AU - Mettes, Pascal

AU - van Gemert, Jan

AU - Smeulders, AWM

PY - 2016

Y1 - 2016

N2 - This method introduces an efficient manner of learning action categories without the need of feature estimation. The approach starts from low-level values, in a similar style to the successful CNN methods. However, rather than extracting general image features, we learn to predict specific video representations from raw video data. The benefit of such an approach is that at the same computational expense it can predict 2D video representations as well as 3D ones, based on motion. The proposed model relies on discriminative Wald-boost, which we enhance to a multiclass formulation for the purpose of learning video representations. The suitability of the proposed approach as well as its time efficiency are tested on the UCF11 action recognition dataset.

AB - This method introduces an efficient manner of learning action categories without the need of feature estimation. The approach starts from low-level values, in a similar style to the successful CNN methods. However, rather than extracting general image features, we learn to predict specific video representations from raw video data. The benefit of such an approach is that at the same computational expense it can predict 2D video representations as well as 3D ones, based on motion. The proposed model relies on discriminative Wald-boost, which we enhance to a multiclass formulation for the purpose of learning video representations. The suitability of the proposed approach as well as its time efficiency are tested on the UCF11 action recognition dataset.

KW - Multiclass Waldboost

KW - video representations

KW - action recognition

KW - feature learning

U2 - 10.1109/ICIP.2016.7532346

DO - 10.1109/ICIP.2016.7532346

M3 - Conference contribution

SN - 978-1-4673-9962-3

SP - 196

EP - 200

BT - 2016 IEEE International Conference on Image Processing (ICIP)

PB - IEEE

CY - Piscataway

Y2 - 25 September 2016 through 28 September 2016

ER -