Page 78 - Kaleidoscope Academic Conference Proceedings 2022

2022 ITU Kaleidoscope Academic Conference

Loading scene, all component datasets are then downloaded. After the successful download, the first AR scene to track the
Theoretically, all components can then be tracked at once. first component is automatically opened. The AssetBundle
But currently, with Vuforia, only one MT can be tracked at is requested from the server and downloaded. The correct
the same time. That is why we created a separate AR scene dataset is automatically created out of the earlier downloaded
for each component and the scenes are tracked successively.
.xml and .dat file. It took 0.66 sec., from entering the AR
If the user selects train number 001, the train consists of the scene to an instantiated MT (line16). Line 17 indicates
locomotive, wagon 1 and wagon2. All three datasets (but not the presentation of the guide view and line 18 a successful
the dataset of wagon 3) are downloaded from the server. In tracking of the locomotive. From showing the guide view
the first AR scene, the locomotive AssetBundle is loaded and until successful tracking, almost a second passed. This value
augmented. In the second AR scene, wagon 2 is augmented heavily depends on the user and in what angle the device
and the AssetBundle is loaded. is held. Successful tracking of Wagon 1 (lines 27 and 28)
took more than 4 sec. When the next scene is opened, the
It is assumed that the locomotive is part of five train MT resets tracking (line 19) and the AssetBundle is unloaded
variants and the AR content for it needs to be updated. (line 20).
Instead of changing five variants, only the one component is
changed. Since all components are remotely stored, the The performance evaluation is given by time but cannot be
application does not even need to be updated with what is classified and compared to other frameworks or applications
better for the user. because there is no reference. For the given use case it was
important to get results that are acceptable for the user and
6. EVALUATION less than one second to download all datasets and about 2
seconds to load and track the MT is in an acceptable range.
This section presents the evaluation of all prototypes as
envisaged in design science research. The evaluation of the Table 1 – Performance evaluation: XCode console
various prototypes with multiple components and the final output
prototype are described below.
line XCode Console output
6.1 Performance 1 14:35:14.39 Vuforia Initialized
2 14:35:15.46 Entering Start scene
3 14:35:29.56 Components selected.
To evaluate the performance major actions were marked in
scripts and printed to the console with a time stamp. These 4 14:35:30.09 Entering Loading Scene
actions include: 5 File saved at: /var/mobile/Containers/Data/Application/Documents/
Vuforia/Lok.xml
• User activities including scene changes 6 File saved at: . . . /Vuforia/Lok.dat
7 File saved at: . . . /Vuforia/Wagon1.xml
• Successful download of files 8 File saved at: . . . /Vuforia/Wagon1.dat
9 File saved at: . . . /Vuforia/Wagon2.xml
• Application requests to the server 10 File saved at: . . . /Vuforia/Wagon2.dat
• Instantiating the empty MT object with the 11 14:35:30.77 Entering AR Scene 1
AssetBundle 12 Requesting bundle at http://www.westfahlsophie.com/Lok-IOS
13 Creating dataset: Vuforia/Lok.xml/Lok
• Target tracking status 14 14:35:30.83 Target: - EMPTY - NO_POSE - NOT_OBSERVED
15 Loaded: ModelTarget (4)(Clone)
• Unload time of an AssetBundle 16 14:35:31.43 Target: Lok NO_POSE - NOT_OBSERVED
17 14:35:31.48 Target: Lok NO_POSE -RECOMM_GUIDANCE
The application was executed on an iOS device and the 18 14:35:32.41 Target: Lok TRACKED - NORMAL
performance of the XCode console output was evaluated. 19 14:35:34.52 Target: Lok NO_POSE - NOT_OBSERVED
The app started on an iPhone X and a speed test indicated a 20 UnloadTime: 7.174000 ms
download rate of 100MB/s. We used a simple virtual server 21 14:35:34.49 Entering AR Scene 2
that is reachable for everyone. The used dataset files had an 22 14:35:34.53 Target: - EMPTY - NO_POSE - NOT_OBSERVED
average size of 83 KB. 23 Requesting bundle at http://www.westfahlsophie.com/Wagon1-IOS
Vuforia/Wagon1.xml/Wagon1
Creating
24
dataset:
25 Loaded: ModelTarget (1)(Clone)
In Table 1, the significantly reduced XCode console output can 26 14:35:34.89 Target: Wagon1 NO_POSE - NOT_OBSERVED
be seen. Line 1 shows that the app has started and Vuforia is 27 14:35:34.95 Target: Wagon1 NO_POSE -RECOMM_GUIDANCE
initialized. Then, the start scene is opened and the variant 28 14:35:38.97 Target: Wagon1 TRACKED - NORMAL
with its components is selected.
6.2 Usability
The Loading scene is entered in second 30.09 and all dataset
files (3 .dat and 3 .xml files) are downloaded to the device Brooke notes that usability as such cannot be clearly defined
within 0.69 sec. The dataset of Wagon 3 was not downloaded, and is always dependent on a context and a user group. To
because it was not selected in the Input scene. still be able to quantify and evaluate usability, he presented
the System Usability Scale (SUS) in 1996 [16]. It includes

– 32 –