ICACT20260046 Question.1
Questioner: zhaozx@dsp.ac.cn 2026-02-10 오후 4:42:27 |
ICACT20260046 Answer.1
Answer by Auhor jianms@nfu.edu.tw 2026-02-10 오후 4:42:27
Chrome  Click!! |
| What unique characteristics of XR (e.g., AR, VR, MR) make it necessary to transmit large amounts of video/image data? How does multi-user interaction in XR further increase the demand for real-time data transmission? |
We would like to use the text content instead of the full image for transmitted data reducing.
Especially, if the image itself is not the only important content for presentation.
In other words, the performance of content exchanging could be enhanced. |
ICACT20260046 Question.2
Questioner: gonghao23@mails.ucas.ac.cn 2026-02-24 오전 7:34:39 |
ICACT20260046 Answer.2
Answer by Auhor jianms@nfu.edu.tw 2026-02-24 오전 7:34:39
Chrome Click!! |
| Accuracy vs. Compression Trade-off: The paper demonstrates that transmitting text descriptions instead of images reduces data to only 1–1.25% of the original size, but how does the system ensure that the regenerated images accurately preserve the semantic intent of the original, especially for complex or emotion-heavy scenes where object attributes alone may be insufficient? |
It is a very good question and issue.
Yes, with less prompts, the content or new image would be more different from the original one.
Hence, it also depends on "how many words to be used and tranferred ?" and "how the same image is required".
In this research, we do not give and estimate the issue yet. |
ICACT20260046 Question.3
Questioner: gonghao23@mails.ucas.ac.cn 2026-02-24 오전 7:37:15 |
ICACT20260046 Answer.3
Answer by Auhor jianms@nfu.edu.tw 2026-02-24 오전 7:37:15
Chrome Click!! |
| Scalability of the Tree Structure: The proposed VLM-based object relation construction uses a tree structure rooted at the most prominent subject — but how does the system handle highly dynamic or crowded scenes with multiple persons of similar size, where defining a single root node and maintaining meaningful inter-object relationships becomes significantly more complex? |
In our research, we could select various definitions.
For example, with the recognized object "human", it will be one root node of the structure.
If multiple people recognized, it means multiple tree structures will be built.
|
