THUMOS Challenge 2014 The International Workshop and Competition on Action Recognition with a Large Number of Classes In conjunction with ECCV’14, September 7, 2014, Zurich, Switzerland http://crcv.ucf.edu/THUMOS14/ **ACTION RECOGNITION IN TEMPORALLY UNTRIMMED VIDEOS** Automatically recognizing and localizing a large number of action categories from videos “in the wild” is of significant importance for video understanding and multimedia event detection. THUMOS workshop and challenge aims at exploring new challenges and approaches for large-scale action recognition with large number of classes from open source videos in a realistic setting. Most existing action recognition datasets are composed of videos that have been manually trimmed to bound the action of interest. This has been identified to be a considerable limitation as it poorly matches how action recognition is applied in practical settings. Therefore, THUMOS 2014 will conduct the challenge on temporally untrimmed videos. The participants may train their methods using trimmed clips but will be required to test their systems on untrimmed data. A new forward-looking dataset composed of the following components will be made available under this challenge: Training Set: over 13,000 temporally trimmed videos. Validation Set: Over 1000 temporally untrimmed videos with temporal annotations of actions. Test Set: Over 1000 videos with withheld ground truth. Spatio-temporal Annotations: Bounding box annotation for 24 actions. All videos are collected from YouTube, and their pre-extracted low-level features (e.g., Dense Trajectory Features, Improved DTF, STIP, etc.) will be made available. We will evaluate the success of the proposed methods based on their performance on the new THUMOS 2014 Dataset in three tasks: Action Classification: this task accepts submissions for whole-clip action classification over 101 action classes. Temporal Action Localization: this task accepts submissions on action recognition and temporal localization on over 24 action classes. Spatio-temporal Action Localization: this task accepts submissions on action recognition and spatio-temporal localization on over 24 action classes. Participants may either submit a notebook paper that briefly describes their system, or a research paper detailing their approach. All of the submission results will be summarized during the workshop and included in the workshop\conference proceedings. Additionally, the top performers will be invited to give oral presentations, with remaining entries encouraged to present their work in the poster session. Important Dates: Development Kit Release (Training+Validation data, low-level features, and evaluation setup): June 2014 Test Data Release: August 3, 2014 Submission Deadline: August 17, 2014 Challenge Results Notifications: August 24, 2014 Publication date: As per conference schedule Workshop date: September 7, 2014 More information available at the workshop website: http://crcv.ucf.edu/THUMOS14/ Program Chairs: Yu-Gang Jiang, Fudan University Jingen Liu, SRI International Amir Roshan Zamir, UCF George Toderici, Google General Chairs: Ivan Laptev, INRIA Mubarak Shah, UCF Rahul Sukthankar, Google Research