ICACT20220027 Question.1
Questioner: jianms@nfu.edu.tw 2022-02-16 오전 11:12:17 |
ICACT20220027 Answer.1
Answer by Auhor 21049228g@connect.polyu.hk 2022-02-16 오전 11:12:17
Chrome  Click!! |
| Since the proposed system is used for detect the real time environment, how about the performance and time delay in detection and tracking? |
Thank you very much for your interest in this article. The experimental results show that the real-time target detection model for the red ball achieves 89.93 FPS, while the target tracking model achieves 50.12 FPS. |
ICACT20220027 Question.10
Questioner: vuvietvu@vnu.edu.vn 2022-02-17 오전 12:13:54 |
ICACT20220027 Answer.10
Answer by Auhor 21049228g@connect.polyu.hk 2022-02-17 오전 12:13:54
Chrome Click!! |
| Thanks for your work and presentation. I have a question that your proposed method currently works on‐chip directly on smart camera or it only processes on the video that capture from the camera. Besides, can you please let me know what kind of camera that you used in your experiments. |
|
ICACT20220027 Question.11
Questioner: vuvietvu@vnu.edu.vn 2022-02-17 오전 12:17:50 |
ICACT20220027 Answer.11
Answer by Auhor 21049228g@connect.polyu.hk 2022-02-17 오전 12:17:50
Chrome Click!! |
| Besides the FPS, have you evaluated your methods in other aspects, for example, the accuracy? Thanks. |
|
ICACT20220027 Question.2
Questioner: tomayoon@ieee.org 2022-02-16 오후 12:04:08 |
ICACT20220027 Answer.2
Answer by Auhor 21049228g@connect.polyu.hk 2022-02-16 오후 12:04:08
Chrome Click!! |
| First of all, thank you for the good research paper and presentation contents. I would like to know more about the research environment used in this study, and also in detail about the equipment & operating systems, development languages and software libraries you used. |
Thank you very much for your interest in this article. This experiment uses the YOLOv3-tiny and KCF models, while the NX development board is the core of the whole embedded terminal deployment. By connecting a camera to capture image information and using TensorRT to accelerate the YOLOv3-tiny model, the detection model is ported to the NX development board and the red ball positioning and size information is output to the LCD through target detection. NX relies on the UART serial port to communicate with the car's motion module to control the car's left/right/back/forward motion. In addition to this, NX requires the installation of common libraries such as OpenCV and JetPack. In this paper, the YOLOv3-tiny target detection model was built using PyTorch, a deep learning framework that supports parallel training on multiple cards, and the CUDA library was used to speed up the computation on the graphics card. The experiment is hosted on Ubuntu 18.04.5 LTS with an Intel Xeon Gold 6149 processor and a NVIDIA RTX 2080Ti graphics card to meet the model training requirements. In order to run the target detection model on the NX platform, the trained model needs to be accelerated by TensorRT. The experiments were conducted using a server to train a PKUAIP dataset. |
ICACT20220027 Question.3
Questioner: franklinlu888@outlook.com 2022-02-16 오전 11:19:37 |
ICACT20220027 Answer.3
Answer by Auhor 21049228g@connect.polyu.hk 2022-02-16 오전 11:19:37
Chrome Click!! |
| Will you tell us what is the major achievement of this paper, and also what is further study issue from this? Thanks in advance!! |
Thank you for your question. In this paper a real time target detection and tracking system has been implemented. There is still much room for research on this system. The following are some directions worth exploring. Firstly, the overall performance of the model can be improved by adding pre-processing such as defogging. Secondly, infrared detection devices can be added to the system to achieve monitoring and tracking of live animals, which can be applied to the monitoring of animal activities in the field. Finally, for the target detection model and tracking algorithm in this paper, further compression can be applied by means of model pruning and quantization. |
ICACT20220027 Question.4
Questioner: noguchi@is.ritsumei.ac.jp 2022-02-16 오전 11:24:39 |
ICACT20220027 Answer.4
Answer by Auhor 21049228g@connect.polyu.hk 2022-02-16 오전 11:24:39
Chrome Click!! |
| Can your proposed system also be applied to recognizing and tracking a vehicle in front? In a system that tracks a target vehicle, could it be used to recognize the vehicle's license plate in front and notify the system when the vehicle in front changes? |
Thank you very much for suggesting a very good research direction, which will also be addressed in our next research extension to extend the functionality of the model to practical application scenarios, applying it to tracking and identifying the license plates of vehicles behind walls, and furthermore having the tracking and identification of traffic signs around the vehicle during its journey. |
ICACT20220027 Question.5
Questioner: ruigelwang@163.com 2022-02-16 오전 11:27:56 |
ICACT20220027 Answer.5
Answer by Auhor 21049228g@connect.polyu.hk 2022-02-16 오전 11:27:56
Chrome Click!! |
| Have you applied this real-time object detection and tracking method to the actual scene, and how accurate is the target detection rate? |
Thank you very much for your interest, we are now working on the application of the model in a variety of realistic scenarios and that accuracy is averaging over 80% and improvements are being made to the performance of the model. |
ICACT20220027 Question.6
Questioner: zakaria.sawado@gmail.com 2022-02-16 오전 11:30:20 |
ICACT20220027 Answer.6
Answer by Auhor 21049228g@connect.polyu.hk 2022-02-16 오전 11:30:20
Chrome Click!! |
| Thank you for the good content of the research paper and the presentation. I would like to know where your research stands in relation to the state of the art. |
Thank you very much for your interest in this article. As mentioned in the article, we have technically implemented a well-performing real-time target detection and tracking system and will be working on the application of the model in various life scenarios. |
ICACT20220027 Question.7
Questioner: ycc.nttu@gmail.com 2022-02-16 오전 11:32:07 |
ICACT20220027 Answer.7
Answer by Auhor 21049228g@connect.polyu.hk 2022-02-16 오전 11:32:07
Chrome Click!! |
| Thank you for your analysis and discussion of this Real-time Object Detection and Tracking System. Do you have any suggestions for future applications with your way? |
Thank you very much for your interest in this article. As mentioned in the article, we have technically implemented a well-performing real-time target detection and tracking system. There is still a lot of room for research on this system. For example, the overall performance of the model could be improved by adding pre-processing (e.g. defogging), or infrared detection devices could be added to the system to enable monitoring and tracking of live animals. |
ICACT20220027 Question.8
Questioner: thanhphuong@lhu.edu.vn 2022-02-16 오전 11:36:28 |
ICACT20220027 Answer.8
Answer by Auhor 21049228g@connect.polyu.hk 2022-02-16 오전 11:36:28
Chrome Click!! |
| Can you tell us why you chose YOLOv3 in your method and what advantages it has. Thank you very much |
Thank you very much for your question, this paper uses the YOLOv3-tiny model. The first stage of the detector covered YOLO v1, v2, v3 and SSD. The main consideration for the One-stage detector was speed, and an improved version of YOLO based on this is called YOLOv3-tiny, which has improved speed and accuracy compared to the previous YOLO series. |
ICACT20220027 Question.9
Questioner: namacabale@gmail.com 2022-02-16 오후 12:11:28 |
ICACT20220027 Answer.9
Answer by Auhor 21049228g@connect.polyu.hk 2022-02-16 오후 12:11:28
Chrome Click!! |
| Dear Author, you have a very good research. I would like to know if the technology today is capable of implementing your proposed scheme? Is there an existing technology almost similar to your proposal that has been deployed in real world environment? Thank you and stay safe.
|
Thank you very much for your interest in this paper's solution, as far as I am aware the technical basis exists to implement it, for example, according to the relevant literature and experiments I have read, existing techniques already enable defogging pre-processing, but so far I have not seen any articles that fully overlap with our direction. All we need to do is to integrate this technology with our model and incorporate this defogging pre-processing technology in the detection and tracking of specific targets in real time. |
