THUMOS: The First International Workshop on Action Recognition with a Large Number of Classes, in conjunction with ICCV '13, Sydney, Australia.
*THUMOS 2014 (featuring TEMPORALLY UNTRIMMED videos)
to be held in conjuction with ECCV 2014*
THUMOS'13 Action Recognition Challenge Overall Results:
Rank | Submission | Overall Accuracy | Split 1 Acc. | Split 2 Acc. | Split 3 Acc. |
1 | ID39_INRIA | 85.900% | 84.734% | 85.862% | 87.105% |
2 | ID40_Florence | 85.708% | 85.319% | 86.642% | 85.164% |
3 | ID35_Canberra | 85.437% | 84.761% | 86.367% | 85.183% |
4 | ID38_CAS_SIAT | 84.164% | 83.515% | 84.607% | 84.368% |
5 | ID25_Nanjing | 83.979% | 83.111% | 84.597% | 84.229% |
6 | ID34_UCF_BoyrazTappen | 82.829% | 82.640% | 83.352% | 82.496% |
7 | ID36_UCSD_MSRA_SJTU | 80.895% | 79.410% | 81.251% | 82.025% |
8 | ID28_USC | 77.360% | 76.154% | 77.704% | 78.222% |
9 | ID31_NII | 73.389% | 71.102% | 73.671% | 75.393% |
10 | ID44-UNITN | 70.504% | 70.446% | 69.797% | 71.270% |
11 | ID42-UEC | 66.261% | 65.157% | 66.726% | 66.899% |
12 | ID26_UMD | 65.948% | 65.218% | 65.385% | 67.240% |
13 | ID47_UNAL | 65.675% | 65.313% | 65.480% | 66.231% |
14 | ID32_Buffalo | 64.296% | 63.405% | 65.365% | 64.118% |
15 | ID29_TNO | 63.457% | 62.007% | 63.461% | 64.904% |
16 | ID37_ECNU | 54.738% | 54.764% | 55.162% | 54.287% |
**Detailed Results and notebook papers available
HERE.**
THUMOS Challenge:
For action recognition to operate in realistic conditions, the vision community needs to make a concerted effort to go beyond datasets with limited number of action classes, such as KTH, Weizmann and IXMAS. The goal of our workshop is to encourage researchers to develop novel methods for action recognition that scale to large numbers of action categories captured in natural settings, both in terms of classification accuracy and computational complexity.
To enable direct comparisons of proposed approaches, we will encourage workshop participants to evaluate their methods on the newly released UCF101 dataset, which is currently the largest action dataset both in terms of number of categories and clips, with more than 13000 clips drawn from 101 action classes.
Since UCF101 dataset contains more than two million frames, we recognize that computing features may itself be a challenge for those workshop participants who lack access to cluster computing resources. Therefore, in order to encourage broad participation, we made available a variety of pre-computed low-level features, such as STIP, SIFT and DTF (Dense Trajectory Feature). While participants are encouraged to employ their own features, the provided features may serve as a useful resource, particularly for computationally-constrained participants. In addition, we will make frame-by-frame bounding box annotations for humans in 24 action classes as well as class-level attribute lists.
Important Dates
submission deadline:
Evaluation results (competition track): November 7, 2013
Papers (research track): November 7, 2013*
*The authors who wish to have their papers published in the ICCV proceedings should submit their papers by the early deadline of September 7.
Competition results announcement: November 28, 2013
Review results and author's notice: October 7, 2013 (early deadline submissions)
Camera ready: October 11, 2013 (early deadline submissions)
Publication date: As per conference schedule
Workshop date: December 7, 2013
[ Downloads | Submission | Call for Papers | Competition Evaluation ]
News:
[11/2013] Detailed challenge results and notebook papers available Here.
[11/2013] Challenge results are now available.
[11/2013] The notebook paper instructions are now available.
[11/2013] The program of the workshop is now available.
[9/2013] The submission website is now open.
[8/2013] We will accept submissions by November 7.
[8/2013] The bounding box annotations of humans are now available for download.
[8/2013] The class-level attributes are now available for download.
[8/2013] The predefined splits for training-testing (recognition) are now available for download.
[7/2013] The low-level features (STIP, DFT) are now available for download.
[7/2013] The workshop benchmark dataset can be downloaded at UCF101.
[7/2013] website launched.
Workshop Organization
General Chairs:
Ivan Laptev, INRIA
Massimo Piccardi, Univ. of Tech., Sydney
Mubarak Shah, UCF
Rahul Sukthankar, Google Research
Yu-Gang Jiang, Fudan University
Jingen Liu, SRI International
Amir Roshan Zamir, UCF
All names ordered alphabetically.
keynote speakers
Jason J. Corso, SUNY at Buffalo, USA
Tal Hassner, Open University, ISrael
Silvio Savarese, Stanford, USA
Cordelia Schmid, INRIA, France
Stan Sclaroff, Boston University, USA
Jianxin Wu, NJU, China
program committee
Saad Ali, SRI International, USA
Marco Bertini, University of Florence, Italy
Cigdem Beyan, University of Edinburgh, UK
Francois Bremond, INRIA, France
Liangliang Cao, IBM T. J. Watson, USA
Jason J. Corso, SUNY at Buffalo, USA
Riad Hammoud, BAE Systems, USA
nazli ikizler-cinbis, Hacettepe Univ., Turkey
Quoc V. Le, Stanford University, USA
Zicheng Liu, MSR, USA
Jiebo Luo, University of Rochester, USA
Greg Mori, SFU, Canada
Ronald Poppe, Univ. of Twente, Netherlands
Michalis Raptis, UCLA, USA
Michael S. Ryoo, NASA JPL, USA
Shin'ichi Satoh, NII, Japan
Silvio Savarese, Stanford, USA
Stan Sclaroff, Boston University, USA
Cees Snoek,Univ. of Amsterdam,Netherlands
Sinisa Todorovic, Oregon State Univ., USA
Yang Wang, University of Manitoba, Canada
Jianxin Wu, NJU, China
Lexing Xie, ANU, Australia
Shuicheng Yan, NUS, Singapore
Alper Yilmaz, OSU, USA
Junsong Yuan, NTU, Singapore
Lihi Zelnik-Manor, Technion, Israel
Data Collection
Khurram Soomro, UCF
sponsors
ICCV International Workshop on Action Recognition with a Large Number of Classes, Sydney, Australia, 2013