Database Reference
In-Depth Information
Table 4.3
A summary of the
subjective survey
Q# Valid result Criteria 1 2 3 4 5 Avg.
1 10 Useful 0 1 2 1 6 4.2
2 10 Satisfied 0 1 1 3 5 4.2
3 10 Satisfied 0 1 1 4 4 4.1
4 10 Satisfied 0 2 2 2 4 3.8
5 9 Useful 0 1 1 3 4 4.11
6 10 Useful 0 1 3 2 4 3.9
7 10 Useful 0 1 1 4 4 4.1
8 10 Useful 0 1 1 4 4 4.1
9 10 Useful 0 1 2 3 4 4.0
A scale of 1-5 is used, with 5 indicating the most use-
ful/satisfied level, 1 indicates the least useful/satisfied level,
and 3 is the neutral
Question 8 and 9 are about the overall usefulness in terms of a recommendation
system and
TapTell
as an application for mobile devices. Most people gave
positive response to the usefulness of this system for both recommendations, as
well as the application in general.
The last question asks a price (in USD) they would be willing to pay at the mobile
market to obtain this application. Eight out of ten people prefer a price less than
$4
99, where two are not willing to pay anything. The remaining two participants
are willing to pay a price above $10.
.
On average, questionnaire participants were satisfied with the
TapTell
system.
Most responses were either 4 or 5 s on the 5-point scale. They also provided
insightful comments such as
Quote 1
“Maybe can cooperate with the fashion industry.”
Quote 2
“This is quick and natural. Better than pre-segmented based method. The segment
results are always confusing.”
4.5
Summary
A contextual-based mobile visual search utilizing the BoW model is used in this
Chapter. A viable application,
TapTell
, is implemented to achieve mobile recogni-
tion and recommendations. Meaningful social tasks and activities are suggested to
users with the assistance of multimedia tools and rich contextual information in the
surroundings. Different gestures have been investigated from tapping the segments,
to drawing the lines of rectangle, to making an “O”-circle via the multi-touch screen.
It is demonstrated that the “O” behavior is the most natural and agreeable user-
mobile interaction. Along with the BoW model, a context-embedded vocabulary
tree for soft weighting is adopted by using both “O” object and its surrounding
image context to achieve mobile visual intents mining. Various weighting schemes
were evaluated with and without GPS conditions, and verified that image context
Search WWH ::
Custom Search