Downloads

The new THUMOS 2014 data can be downloaded using the following links.
The details of the competition tasks, evaluation metrics, dataset, submission format, etc. can be found in the Evaluation Setup Document (updated 08/20/2014).

Password:

Please fill this form in order to receive the password required for unzipping some of the shared data.

Dataset, Annotations, Development Kit:

Training Data (13320 trimmed videos) -- each includes one action:

UCF101 videos (zipped folder): [Download] (updated Jul. 27, 2014)

UCF101 videos (individual files): [Link] (updated Jul. 27, 2014)

Description of UCF101: [Link]

Pre-computed low-level Features (Improved DTF [1]): [Download] (updated Jul. 27, 2014)

List of action classes and their numerical index: [Download]

Background Data (2500 untrimmed videos) -- each guaranteed to not include any instance of the 101 actions:

Videos (zipped folder): [Download] (updated Jul. 31, 2014)

Videos (individual files): [Link] (updated Jul. 31, 2014)

Metadata and annotations (primary action class for each video): [Download] (updated Jul. 31, 2014)

Pre-computed low-level Features (Improved DTF [1]): [Download] (updated Jul. 31, 2014)

Validation Data (1010 untrimmed videos) -- each includes one/multiple instances of one/multiple actions:

Videos (zipped folder): [Download] (updated Jul. 30, 2014 - note)

Videos (individual files): [Link] (updated Jul. 15, 2014)

Metadata and class-level annotations (action classes in each video): [Download] (updated Jul. 30, 2014)

Pre-computed low-level Features (Improved DTF [1]): [Download] (updated Jul. 30, 2014)

Temporal annotations of actions (videos of 20 classes): [Download] (updated Jul. 30, 2014)

Development Kit (evaluation code & additional software): [Link] (updated Oct. 5, 2015)

Additional Data:

Codebook used for quantizing the Improved DTF features: [Download] (updated Jul. 20, 2014)

Class-level attributes: [Download]

Bounding box annotations of humans (24 classes of UCF101): [Download]

Sample submission files: [Temporal Action Detection],[Action Recognition] (updated Jul. 30, 2014)

Test Data (1574 untrimmed videos) -- ground truth made available after the competition:

Videos (zipped folder): [Download] (updated Aug. 18, 2014)

Videos (individual files): [Link] (updated Aug. 18, 2014)

Pre-computed low-level Features (Improved DTF [1]): [Download] (updated Aug. 18, 2014)

Metadata and class-level annotations (action classes in each video): [Download] (updated Oct. 3, 2014)

Temporal annotations of actions (videos of 20 classes): [Download] (updated Oct. 3, 2014)

If you make use of the data and resources shared for the competition, e.g., extracted low-level features or the attribute lexicon, or want to cite the THUMOS challenge, please use the following reference:

@misc{THUMOS14,
   author = "Jiang, Y.-G. and Liu, J. and Roshan Zamir, A. and Toderici, G. and Laptev,
   I. and Shah, M. and Sukthankar, R.",
   title = "{THUMOS} Challenge: Action Recognition with a Large
   Number of Classes",
   howpublished = "\url{http://crcv.ucf.edu/THUMOS14/}",
   Year = {2014}}

UCF101 Dataset can be cited as:

@inproceedings{UCF101,
   author = {Soomro, K. and Roshan Zamir, A. and Shah, M.},
   booktitle = {CRCV-TR-12-01},
   title = {{UCF101}: A Dataset of 101 Human Actions Classes From
   Videos in The Wild},
   year = {2012}}

Participants are strongly encouraged to read the Evaluation Setup Document for the details of the competition.

We truly apprecaite the following who spent many hours performing annotation/verification for THUMOS:
INRIA: Anastasia Syromyatnikova, Ivan Laptev.
UCF: Ann Dang, Cory Kinberger, Lucas Pasqualin, Jose Rodriguez, Alejandro Torroella, Fatemeh Yazdian.
FUDAN: Qi Dai, Xue Guo, Huijuan Liu, Chao Ma, Xiaoxin Qiu, Guorui Sun, Zuxuan Wu, Nina Xiang, Yangjin Yao, Yiye Zhu.


[1] Heng Wang and Cordelia Schmid, "Action Recognition with Improved Trajectories," in 2013 IEEE International Conference on Computer Vision (ICCV).