Network-based visualization of very large 3D landscape and city models
Jérôme Royan, Patrick Gioia, Romain Cavagna, Christian Bouville
This page contains complementary documents related to the paper
untitled "Network-based visualization of very large 3D landscape and
city models".
Adaptive view-dependent streaming
A demo video
showing navigation through 3D lanscapes and city models. It
demonstrates the efficiency of the wavelet-based terrain model and the
PBTree LOD model for flyover navigation. An overview of the LOD techniques can be found here.
On-demand streaming technique with the LODDT
The following two videos has been obtained with our real time
simulation platform for network-based visualization. In each of these
simulations, all clients follow a predefined path in the 3D
environment and we assess the performances of the streaming techniques
by measuring in real time the accuracy each client.
Accuracy (ACC) represents the discrepancy at time t between the
required data for the current VP viewpoint, and the received data that
can be used at this position. In our case, as we use a hierarchical
model for the objects geometry, peer accuracy at time t is measured as
the proportion of missing nodes (NMISS) in the PBTree with
respect to the number of required nodes (NREQ) at the current
viewpoint: ACC = 1 - NMISS/NREQ.
This video shows the
efficiency the on-demand streaming technique base on the LODDT compared
to the former one-step method. Boths are simulated concurrently. The
buildings loaded with one-step method are in colored in green wheras those loaded with the LODDT-based method are drawn in grey.
The image below is a snapshot of the video. It can be seen that for the same simulation time, there are much more
buildings loaded with LODDT-based method than with one-step method.
Comparison of client-server and peer-to-peer architecture : simulation of abrupt change in network load
This other video shows
simulation results obtained with our peer-to-peer overlay in comparison
with a client-server architecture. It simulates an abrupt change in
system load when 10 peers arrive in the same time and start to navigate
in the environment. It shows peers moving in the 3D environment and the
min, max and average accuracy values are drawn on a graph located in
the right hand side of the image. Peers connectivity are represented by
connection lines and we can see how it evolves as peers move in the
enviromment. The first part of video shows the results obtained with a
classic client-server architecture and the second part shows the
performances brought by the peer-to-peer overlay. It can be seen that
the the maximum level of accuracy (ACC = 1) is reached more than twice
faster with the peer-to-peer overlay.
The picture below is a snapshot of the video in the client-server case
(upper image) the peer-to-peer case (lower image). It can be seen
that the average accuracy curve (in brown in the graph on the
right) is following a much steeper slope in the peer-to-peer case,
which demonstrates the fast self-organisation property of the
peer-to-peer overlay.
The
client-server case
The peer-to-peer case