TRECVID 2014 Evaluation for Surveillance Event Detection

TRECVID 2014 Evaluation for Surveillance Event Detection

The Interactive Surveillance Event Detection (SED) evaluation is a track within the TRECVID Evaluation. The goal of the 2014 Interactive SED evaluation track is to support the development of interactive (i.e., human in the loop) technologies to detect visual events (people engaged in particular activities) in a large collection of streaming video data. The interactive aspect of SED was added in 2012. The evaluation will be implemented using an 11 hour subset of the multi-camera airport surveillance domain evaluation data collected by the Home Office Scientific Development Branch (HOSDB).

The 2014 SED evaluation is the 6th evaluation of event detection technologies preceded by the 2008, 2009, 2010, 2011, 2012 and 2013 evaluations. The interactive evaluation methodology is similar to previous evaluation, i.e., systems report the same information and are scored with the same metrics, however, unlike previous evaluations the data set size will be smaller and each search for an event by a searcher can take no more than 25 elapsed minutes (wall clock time), measured from the time the searcher is given the event to look for until the time the result set is considered final. The system may use an unlimited amount of computional resources prior to beginning the interactive search.

We have a mailing list for announcements/discussion -- contact martial.michel [at] nist.gov (martial[dot]michel[at]nist[dot]gov) to subscribe.

Evaluation Planning

The following resources are available:

• Evaluation Plans: Read the TRECVID SED 2014 Evaluation Plan-V01
• Event Definitions:
  • An annotation guidelines document TRECVid09_Event_Annotation_Guidelines_v1.0.pdf is available.
  • The following paper EventDet-ViPER-format-v4.pdf documents the XML Schema Definition (XSD) proposed for a ViPER XML Representation of Events.
  • The following TrecVid08_XSD_v2.tar includes the XSD and a sample XML file for events.
• Experiment Control Files:
  • The following TRECVid08_ecf_v2.tar includes the XSD and a sample XML file for Experiment Control Files (ECFs).

Software

Evaluation submissions will be scored using the Framework for Detection Evaluations (F4DE) toolkit downloadable from the MIG Tools Web Page. You need at least F4DE 3.1.1 (released July 3rd 2014) to use the SED14 SubmissionChecker tool.

Dry Run

The dry run period for the 2014 TRECVID Interactive Surveillance Event Detection Evaluation will run until July 16th 2014. Dry run submissions will be accepted at any time during that period. The procedure for participating is as follows:

• Sign up as a TRECVID Participant as specified on the TRECVID web site.
• Obtain the data sets by completing the licensing agreement as specifiec below. The dry run will use part of the development corpus.
• Download the Experiment Control File (ECF) [same as 2009, 2010, 2011, 2012 and 2013]: expt_2009_retroED_DEV09_ENG_s-camera_NIST_2.xml
• Run experimental system(s) on the video intervals specified in the ECF.
• Package system outputs in ViPER XML format and make sure they validate against the validation script (that can be found in F4DE -- see evaluation plan for details)
• Prepare a tar/gzip of all outputs as described in the evaluation plan and submit the archive via FTP. Make sure submissions validate using the TV14SED-SubmissionChecker tool. Note: this has not yet been released.
• Notify martial.michel [at] nist.gov (martial[dot]michel[at]nist[dot]gov). Make sure to include a contact point's information (name, email address, location, site long and short name, and possibly a phone number for us to contact you in case of problems), as well as a possible alternate contact person.
• Submissions are due 11:59PM Eastern Time (UTC-4), July 16th, 2014 (extended from June 30th)
• Results of the dry run will be returned by July 20th, 2014 (approx.).

Formal Evaluation

Submissions for the formal evaluation are due July 23rd, 2014. The procedure for participating is as follows:

• If possible: Complete the dry run as specified above.
• [UPDATED] Download the Experiment Control Files (ECF): There is only one ECF file defined in the eval plan for the interactive event detection and the retrospective event detection: expt_2014_allED_EVAL14_ENG_s-camera_NIST_1.xml (different from previous years' ECF files in that 9 of the 11 hours were not evaluated in the past). Note: the reference data for the Formal Evaluation is being bootstrapped and will evolve as part of the evaluation process.
• Run experimental system(s) on the video intervals specified in the ECF.
• Package system outputs in ViPER XML format and make sure they validate against the validation script (that can be found in F4DE -- see evaluation plan for details).
• Prepare a tar/gzip of all outputs as described in the evaluation plan and submit the archive via FTP. Make sure submissions validate using the TV14SED-SubmissionChecker tool.
• Notify martial.michel [at] nist.gov (martial[dot]michel[at]nist[dot]gov). Make sure to include a contact point's information (name, email address, location, site long and short name, and possibly a phone number for us to contact you in case of problems), as well as a possible alternate contact person.
• Submissions are due 11:59AM Eastern Daylight Time (UTC-4), July 23rd, 2014.
• Results of the formal evaluation will be returned by July 30th, 2014 (approx.).

Data Resources

• The data resources provided for 2014 are the same as 2011.
• For more information, see: Licensing/Participation Agreements