Now 3689 visitors
Today:0 Yesterday:0
Total: 139713 215S 51P 5R
2026-10-21, Week 43
Member Login
TACT Journal Homepage
Call for Paper
Committee
Find My Paper
Author Homepage
Statistics & History
Welcome Message
Paper Procedure
FAQ
Paper Archives
Outstanding Papers
Program & Proceedings
Presentation Platform
Hotel & Travel Info
Photo Gallery
Scheduler Login
Seminar
Archives Login
Sponsors




















Paper Number Author:
Paper Title
Keyword
Q&A Number **Select your question number !!
Questioner eMail:
Question
Answer
by
Author
Save Q&A

* Edit or answer any Q&A by selecting Q&A number Hyper Link below + Write button (Save)
ICACT20230038 Question.1
Questioner: p2109422@mpu.edu.mo    2023-02-13 오후 1:14:09
ICACT20230038 Answer.1
Answer by Auhor wxtdsg@mail.ustc.edu.cn   2023-02-13 오후 1:14:09
Chrome Click!!
What are the current challenges and limitations of multispectral detection and how can dynamic neural network help to address these challenges? Common Multispectral Detection Algorithm use two-stream model, which has large amount of calculation. Dynamic neural network can reduce the computation by skipping the layers or pruning the channels without losing much accuracy.
ICACT20230038 Question.2
Questioner: p2109422@mpu.edu.mo    2023-02-13 오후 1:20:03
ICACT20230038 Answer.2
Answer by Auhor wxtdsg@mail.ustc.edu.cn   2023-02-13 오후 1:20:03
Chrome Click!!
In your neural network model, how much is the difference in latency with and without an FPGA accelerator? General hardware accelerators are difficult to benefit from dynamic architectures, due to the significant irregularities introduced by the redundancy elimination at runtime. Our FPGA accelerator can efficiently process the optimized model and reduce the lantency from 10.6ms to 3.9ms.
ICACT20230038 Question.3
Questioner: p2109422@mpu.edu.mo    2023-02-13 오후 1:21:48
ICACT20230038 Answer.3
Answer by Auhor wxtdsg@mail.ustc.edu.cn   2023-02-13 오후 1:21:48
Chrome Click!!
Will the use of FPGA accelerators affect the accuracy of dynamic neural network models? Yes. But the affect is small. The mAP decrease by 0.3%.
ICACT20230038 Question.4
Questioner: mjs9908@g.skku.edu    2023-02-21 오후 4:42:02
ICACT20230038 Answer.4
Answer by Auhor wxtdsg@mail.ustc.edu.cn   2023-02-21 오후 4:42:02
Chrome Click!!
Hi, thank you for your presentation What is the difference between MobilleNetv2 and Dynamic MobileNetV2 ? We insert the layer gate and channel gate (shown in Fig.2) in the MobileNetV2, so that the layer skipping and channel pruning can be implemented.
ICACT20230038 Question.5
Questioner: 21m107fl@stu.teikyo-u.ac.jp    2023-02-22 오후 1:35:27
ICACT20230038 Answer.5
Answer by Auhor wxtdsg@mail.ustc.edu.cn   2023-02-22 오후 1:35:27
Chrome Click!!
Thank you for your presentation. Regarding your proposed method, what do you think can be improved in future work? The current design requires many manual operations, and the process is complicated. I hope it can be simplified in the future.