List of items (comments, suggestions, questions, etc.) that come up in
the TRECVID 2014 workshop to be used as input in planning TRECVID 2015
(no decisions implied or expressed). Each has an implied question mark
at the end.

SIN
- more runs
- simplified master shot reference
- other data sets

LOC
- offer smaller localization test set to attract more participants
- publicize localization task more
- use of NIST LOC judgments from 2012/13 as training data
- offer 2015 SIN run (system output) for participants interested only in LOC

INS
- drop training conditions A-D (how many images used); assume all used 
- keep training condition E (video vs no video distinction)
- give out ID of the shot from which each image example came
- encode the "a/an vs this/these" distinction as xml attribute, not in tex
- add person + known place X topics
- find ways to emphasize both speed and effectiveness: efficiency

MED
- Revise annotations within the Omnibus data distributions (teams submit
  to NIST)
- Investigate limited resource evaluation conditions
- Investigate release of additional development resources

MER
- Revise MER task definition/evaluation protocol (eliminate
  ambiguity of "key" and focus recounting quality judging bette,...)

SED
- investigate UCF101
- create a small “entry” dataset (2h) from 9Way data
- remove ‘camera4’ from used cameras
- ask participants to rerun algorithms on SED13 data
- send to participants “SED13Comm” (portion of data common from SED13 and SED14) results
- follow up with Brno re:bounding box for video
- compute instance per hour on per camera per event analysis
- go to CVPR / ECCV / ICCV and present at a special session ?


GENERAL
- provide list of links to commonly used tools
- advertise more at conferences