search

YongChen-Exact / TC-CoNet

[Computers in Biology and Medicine - 2023] This is an official PyTorch implementation for Collaborative networks of Transformers and Convolutional neural networks are powerful and versatile learners for accurate 3D medical image segmentation
https://doi.org/10.1016/j.compbiomed.2023.107228
MIT License
7 stars 0 forks source link

attention visualization #1

Open Yffy123456 opened 8 months ago

Yffy123456 commented 8 months ago

Can I ask you for advice on how to write an attention visualization in a dissertation?

YongChen-Exact commented 8 months ago

CAM is a common method of attention visualization, which can be achieved by referring to the following methods: https://github.com/PerXeptron/CAM-Localization.

Yffy123456 commented 8 months ago
捕获

But what I would like to ask is in your paper is feature map visualization? Is the feature map visualization different for CAM, right? Looking forward to hearing from you!

YongChen-Exact commented 8 months ago

Attention heatmaps are used to emphasize which regions are important for the model to make pixel level predictions, while feature maps reflect the model's hierarchical understanding of input images. These are not the same. Fig.5 in the paper is an attention heatmap.

YongChen-Exact commented 6 months ago

Sorry, due to the busy schedule during the Spring Festival, I don't have time to reply to your message. Heat maps and feature maps are different. A heatmap represents the target of network attention. Feature maps are the features learned by a network. You can conduct detailed research on the specific definition.

CY @.***

原始邮件

发件人："Yffy123456"< @.*** >;

发件时间：2024/1/6 15:05

收件人："YongChen-Exact/TC-CoNet"< @.*** >;

抄送人："YongChen-Exact"< @. >;"Comment"< @. >;

主题：Re: [YongChen-Exact/TC-CoNet] attention visualization (Issue #1)

default.PNG (view on web) But what I would like to ask is in your paper is feature map visualization? Is the feature map visualization different for CAM, right? Looking forward to hearing from you!

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>