DASH-3D

In the previous episodes
  • DASH-3D, ACMMM 18
  • 3D Bookmarks in DASH-3D, ACMMM 19 demo track
  • Improve 3D streaming with precomputations on 3D bookmarks (not published)
Things to do for the project (deadline : March, extension : September)

Axel's toughts in italic

  • WP1 : Streaming adaptation for mobile (done)
  • WP2 : Interactive application for content preparation (almost done)
  • WP3 : Dynamic NVEs (too difficult, skippable)
  • WP3bis : Semantic compression (most interesting)

Latest user study : mobile / bookmarks

  • Tutorial
  • Extended tutorial
    • Free navigation without bookmarks, then with bookmarks
    • 1 minute each, no randomization
  • Bookmark tour
    • users need to click on the highlighted bookmark
    • Default policy vs Bookmark enhanced policy
    • 1 minute each, randomized, questionnaire
  • Free navigation
    • 1 of [default policy, optimized policy]
    • 1 minute, randomized

Results

  • 18 participants (15 males, 3 females, average age = 20.7, std = 0.53)
  • 10 preferred the optimized policy, 4 did not perceive any difference, 4 preferred greedy policy
Bookmark tour

h:575 center

Bookmark tour

h:575 center

Free navigation

h:575 center

Free navigation

h:575 center

Number of clicks on bookmarks

Policy Average
Greedy 4 1 1 1 3 3 1 7 6 3
Bookmark 3 5 2 5 10 7 6 4 6 5.33

p-value for statistical significance : 0.06

About the semantic compression

Original text about semantic as written in the project

Many compression techniques for 3D compression exist, but we need to adapt them in the context of NVE navigation. Most 3D scenes being very heterogeneous, they contain very high resolution geometries (tens or hundreds thousands faces) laying next to object with significantly lower quality. Obviously, compression techniques need to be adapted. Similar work is also required for textures that show the same kind of heterogeneity. Integration and selection of content of different nature, and more generally, heterogeneous content will be required.

About the semantic compression

  • um, we don't do compression right now
  • I don't know what to say
  • My initial idea was to start analyzing textures and materials hoping to find things