https://github.com/capsule2077/cam-visualization

可视化技术——CAM

https://github.com/capsule2077/cam-visualization

Science Score: 10.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • CITATION.cff file
  • codemeta.json file
  • .zenodo.json file
  • DOI references
  • Academic publication links
    Links to: arxiv.org
  • Academic email domains
  • Institutional organization owner
  • JOSS paper metadata
  • Scientific vocabulary similarity
    Low similarity (1.6%) to scientific vocabulary
Last synced: 10 months ago · JSON representation

Repository

可视化技术——CAM

Basic Info
  • Host: GitHub
  • Owner: capsule2077
  • Language: Python
  • Default Branch: main
  • Size: 299 KB
Statistics
  • Stars: 0
  • Watchers: 1
  • Forks: 0
  • Open Issues: 0
  • Releases: 0
Created almost 3 years ago · Last pushed almost 3 years ago
Metadata Files
Readme

README.md

CAM-Visualization

可视化技术——CAM
CAM论文:Learning Deep Features for Discriminative Localization
GAP论文:Network In Network
官方仓库参考: https://github.com/zhoubolei/CAM

alt text

说明

原论文的代码比较老了,用的pytorch版本比较旧,所以重写了一遍,用了更新的特征提取方式。

CAM_VGG.py使用了VGG16,修改了原始网络的平均池化层和分类层。使用预训练的权重在MINI-ImageNet数据集上做100分类的微调。MINI-ImageNet数据集可以参考:blog. 只微调全连接层,其余层保持不变。权重文件:VGG16_100.pth,提取码:3v0i

CAM_ResNet.py使用了ResNet18,ResNet本身就是GAP+FC,所以不需要修改网络,直接使用官方的预训练权重即可。

class_label分别包含了100类(label_100.json)和1000类的标签文件(label.json

Owner

  • Name: liulongfei
  • Login: capsule2077
  • Kind: user
  • Location: 广西
  • Company: 桂林电子科技大学

GitHub Events

Total
Last Year