Now 6867 visitors
Today:0 Yesterday:0
Total: 154641
216S 51P 5R
2026-10-22, 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 |
| Author Homepage |
| - Paper Procedure |
| - Journal Procedure |
| - Presentation Tips |
|
| Program & Proceedings |
| Program with Papers |
| Plenary Session |
| Tutorial Session |
|
| Presentation Platform |
| Hotel & Travel Info |
| About Korea |
| Accommodation |
| Transportation |
| VISA |
| Other Infomation |
|
| Photo Gallery |
| Scheduler Login |
| Seminar |
| Archives Login |
|
| Sponsors |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| Save Q&A |
 |
* Edit or answer any Q&A by selecting Q&A number Hyper Link below + Write button (Save)
|
ICACT20240312 Question.1
Questioner: sikandarshigri77@gmail.com 2024-02-05 오후 9:45:15 |
ICACT20240312 Answer.1
Answer by Auhor rrmurali.199@gmail.com 2024-02-05 오후 9:45:15
Chrome  Click!! |
| You have used Eigen CAM for explainability. Did you try GradCAM as well? Is there any specific reason for using Eigen CAM? |
Thanks for your valuable question.
Though both Eigen CAM and GradCAM are popular techniques for visualizing and interpreting CNNs by highlighting important regions of an image, Eigen CAM emphasizes the importance of eigenvalues in the feature maps of CNNs, providing insights into how different parts of the image contribute to the final classification decision. And the GradCAM, on the other hand, computes the gradient of the target class with respect to the final convolutional layer to highlight important regions.
As we tried both CAM techniques, the final choice was made based on the specific requirements of our main task which is to detect the rip currents, and the architecture of the model we used YOLO v8. Also Eigen CAM showed the superiority in terms of performance, ease of implementation, and primarily the compatibility with the architecture of the YOLO v8 model. |
ICACT20240312 Question.2
Questioner: shigriyaseen@gmail.com 2024-02-05 오후 10:39:33 |
ICACT20240312 Answer.2
Answer by Auhor rrmurali.199@gmail.com 2024-02-05 오후 10:39:33
Chrome Click!! |
| Were there any limitations or challenges encountered during the study, and how were they addressed? |
Thanks for your interest in our research work.
Yes, we encountered several challenges during the research process. The first challenge was data collection. Obtaining labeled data for training and testing rip current detection algorithms is challenging due to the scarcity of labeled datasets. There we did research and found a novel rip current detection dataset from AI Hub site which we have used in our study.
The next challenge was due to the complex nature of Rip Current dynamics, making their detection challenging. Hence we used multiple object detection models and compared the results to find the better detection model.
Then Interpretability and Explainability of rip current detection models. The interpretability is important for gaining insights into the factors contributing to rip current formation. Using explainable AI techniques, such as Eigen CAM helped to interpret model decisions and improve model transparency. |
| |
