Organisation internationale de normalisation

7.2Compact Descriptors for Visual Search (CDVS)

Images are foreseen as input for visual search. The temporal extension of images also known as video will enable life interaction. ETRI, TELECOM SudParis, Telecom Italia, Stanford University, and University of Peking contributed databases with images and videos as well as annotations to MPEG for use in this potential standardization activity. Additional input on destructors, high quality reference and low-quality test data is required. The annotations of the databases need to be completed and unified. Furthermore, the software and logistics for a CfP need to built. A CfP will only be meaningful if the before mentioned issues get resolved.

The adhoc group N11841 AHG on Compact Descriptors for Visual Search will work on these issues.
A draft CfP N11875 Draft CfP for Compact Descriptors for Visual Search was made publicly available in order to alert the community to the intentions of MPEG.
In order to prepare the CfP and evaluation, MPEG issued the documents N11873 Workplan for CDVS datasets and evaluation framework and N11874 Draft Evaluation Framework for Compact Descriptors for Visual Search.
Given the open issues, the earliest time for issuing a call will be the 96^th meeting with an evaluation planned prior to the 98^th meeting which is one meeting later than originally anticipated.
A liaison statement N11878 Liaison Letter Template Draft Call for Proposals on CDVS will be sent to several interested organisations.