Database Reference
In-Depth Information
8.5
Video Re-ranking on the Social P2P Network
The video re-ranking method, presented in Chap.
3
, is suitable for the implemen-
tation of a video search engine on a P2P network since there is no requirement
to conduct relevance feedback from users. In this application, a video database in
a peer is indexed by the template-frequency model (TFM) and organized by the
3-layer cosine network. The video retrieval is performed by pseudo-RF through
forward-backward signal propagation between peers. This process does not require
bandwidth to actually transfer video files over the network during the RF learning
process. The video re-ranking system is explained in the following sections.
Fig. 8.16
(
a
-
c
) Retrieval results obtained by user-controlled RF on the centralized database; (
a
)
at Iteration
=
0, the retrieval precision was 0.55; (
b
) at Iteration
=
1, the retrieval precision was
0.70; (
c
)atIteration
=
5, the retrieval precision was 0.80, and (
d
) in the retrieval result obtained
with a P2P retrieval system with pseudo-RF, the retrieval precision was 0.95
Search WWH ::
Custom Search